![1
DOI: 10.1201/9781003371380-1
Chapter 1
Skin Cancer
Classification Using
Image Processing with
Machine Learning
Techniques
NirmalaV., Shashank H. S., Manoj M. M., Satish Royal G.,
and Premaladha J.
School of Computing, SASTRA Deemed to be University, Thanjavur, India
1.1 Introduction
Image classification modalities play a significant role in the health sector. Early diag-
nosis of fatal diseases using various imaging techniques [1] has positively impacted
people’s lives. Our work describes classification of skin cancer images using deep
learning techniques [2]. Skin cancer attacks surrounding cells, resulting in the devel-
opment of a mole on the external layer of the skin that can be categorized as malig-
nant or benign. Many solutions using neural network architectures for diagnosis
of the early stages of skin cancer have been proposed. The classification [3] metrics
used include support vector machine (SVM), relevant vector machine (RVM), and
neural network architectures. These machine learning algorithms pose several con-
straints for input data distribution, such as noise-
free or high-
contrast images, but
these constraints do not apply to the skin cancer classification problem. Instead, it](https://image.slidesharecdn.com/26194005-250514134919-aacbaa93/85/Intelligent-Data-Analytics-Iot-And-Blockchain-Bashir-Alam-Mansaf-Alam-28-320.jpg)
![2 ◾ Intelligent Data Analytics, IoT, and Blockchain
is colour, texture, and structural features that play an essential role in skin cancer
classification. Traditional parametric approaches cannot be used for skin cancer clas-
sification problems since skin lesions have different patterns. Hence deep learning
techniques are used.
The automatic classification process [4] includes preprocessing, feature extrac-
tion, segmentation, and classification, resulting in a handcrafted feature set. However,
since lesions have visual resemblance and are highly correlated due to their colour,
texture, and shape, handcrafted feature extraction is not appropriate for skin cancer
classification. The deep learning approach is therefore preferred. We can feed the
images directly to the model, removing the need for any preprocessing [5] to be
implemented before passing the image to the model. Neural network models are very
effective in extracting specific features from the image. Even though deep learning
models are efficient for classifying skin cancer, the various elements present in skin
lesion images make identifying skin cancer [6] challenging for the following reasons:
▪ The ISIC 2016 and ISIC 2017 skin cancer datasets are highly imbalanced,
with many benign samples.
▪ Many skin lesion images are highly similar, and classifying [7] them into
benign and malignant images is challenging.
The novel modified LCNet model is designed for model training for boosted clas-
sification results even for the less accurate lesions of human skin. An optimization
algorithm is improved with the repeated blocks of batch normalization.
The remainder of this chapter is organized as follows. Section 1.2 describes sig-
nificant earlier work in this area. Section 1.3 explains our research in detail, the data-
sets involved and the model architecture. Sections 1.4 and 1.5 present our results
with their comparative analysis and conclusion.
1.2 Related Works
Skin cancer is a deadly disease that can affect nearby cells of the body. Early detection
and diagnosis is important [8]. Initially, handcrafted feature-
based approaches were
used on dermoscopic images. However, since there is a high correlation between
skin texture and colour in skin images, such approaches are not regarded as suitable
for skin classification problems [9]. As preprocessing operations are unnecessary,
deep convolutional neural networks (DCNNs) have proved helpful. The first time
DCNN was applied to skin cancer images [10] used 129,450 skin disease images
to classify 2032 diseases. The researcher designed a deep learning framework with
two fully convolutional residual networks, one to produce the segmentation result
and the other to produce a coarse classification result. A lesion index calculation
[11] unit was introduced to produce a heat map and refine the coarse classification
results. Iqbal et al. [12] describe the contribution of each pixel towards the classifica-
tion of another model, a convolution model consisting of multiple layers used for](https://image.slidesharecdn.com/26194005-250514134919-aacbaa93/85/Intelligent-Data-Analytics-Iot-And-Blockchain-Bashir-Alam-Mansaf-Alam-29-320.jpg)
![Skin Cancer Classification ◾ 3
multi-
class classification. It had 68 convolutional layers, passing the features from
top to bottom. Zhang et al. proposed an attention residual neural network [13] that
consisted of multiple ARL blocks, which was further followed by global average
pooling and classification layers.
To improve classification efficiency, ensembles of CNNs, consisting of outputs
from the layers of four different models, Google Net, AlexNet, VGG and ResNet,
were created by Barata et al. [14]. Another approach using multiple imaging modali-
ties was also proposed to increase the modularity of a self-
supervised topology clus-
tering network that could classify the unlabeled data without needing class-
based
information. The model learnt features at different levels of variations, such as the
illumination, the background and the point of view.
Some models applied transfer learning [15] using pre-
trained models for dermo-
scopic classification. Nevertheless, a small dataset with high accuracy does not fit all
scenarios, especially with medical images, since each piece of information is highly
sensitive in the diagnosis. Hyperparameter tuning was performed to achieve better
results. Gessert et al. developed an ensemble model from Efficient Nets, SENet, and
ResNet WSL which was used to perform a multi-
class classification task [16] on the
ISIC 2019 dataset. A cropping strategy was implemented to deal with different input
resolutions. With the earlier reports for melanoma classification, many researchers
carried out different trials. They achieved reasonable accuracy, and some of the results
are inspiring. However, the status of early diagnosis of melanoma skin cancer is not
generalized and, as we have seen in the Introduction, there are two major problems.
1.3
Materials and Methods
1.3.1 Dataset
The skin cancer images were obtained from the ISIC 2016 [17], ISIC 2017 [18]
and ISIC 2020 [19] challenges. Since the skin samples were highly imbalanced, they
were augmented by datasets from the ISIC archive. The ISIC challenge provides
two datasets – training and testing. We divided our model learning and estimation
process into three parts: training, validation and testing. We chose 20% for the vali-
dation process. Since the given skin cancer samples were highly imbalanced, data
augmentation was carried out on classes with fewer samples to prevent the deteriora-
tion of the model learning process.
Data were classified into two classes: MEL (malignant) and BEN (benign). Other
lesion types, such as seborrheic keratosis and nevus, are also considered benign. The
total number of training samples and validation samples is shown in Table 1.1. The
data distribution over ISIC 2016, 2017 and 2020 datasets is shown in Figure 1.1.](https://image.slidesharecdn.com/26194005-250514134919-aacbaa93/85/Intelligent-Data-Analytics-Iot-And-Blockchain-Bashir-Alam-Mansaf-Alam-30-320.jpg)
![Skin Cancer Classification ◾ 5
1) The proposed model uses a DCNN for classifying images into benign and
malignant. The model consists of multiple blocks bonded together to facili-
tate the processing of many features in the convolutional neural network
architecture.
2) Each block consists of varying parameters having different values. Parameters
include stride, number of kernels, and kernel size.
3) The model consists of 11 blocks, each having its sequence of operations per-
formed over the image.
1.3.2 Preprocessing Operations
For the given data sets, preprocessing was carried out to make them suitable for
passing through the model. The images were normalized to make the computations
effective. The normalization of images was carried out using pixel normalization.
The pixel values were scaled to 0–1. Normalization is essential because it ensures that
each input parameter has a similar data distribution.
Furthermore, data augmentation – rotation, shifting, flipping, and scaling – was
carried out since the image samples were highly imbalanced [20]. A random rota-
tion of 0° to 90° was applied to the image. The image was shifted by 10% of the
entire width and height. Horizontal and vertical flipping was carried out on all the
images. These augmentation operations [21] were applied only to the training and
validation sets. Sample augmentation operations are shown in Figure 1.2. Since the
model accepts a (128, 128, 3) image, all the images in the dataset were resized [22]
to (128, 128, 3).
For the model, we stored all the images in HDF5 format and organized them
into folders depending on the type and category of the image [23].
1.3.3 LCNet Architecture
The proposed DCNN model, Lesion Classification Network (LCNet), is formu-
lated using 11 blocks, as shown in Figure 1.3. The model also has the following
specification:
▪ Block 4 and 5 – repeated twice
▪ Block 7 and 8 – repeated 4 times
▪ Blocks 10 and 1 – repeated twice
The network accepts a (128 × 128 × 3) image as an input, after which a convolu-
tion operation is performed over the image using a (3 × 3) kernel having a stride of
2 to learn 8 features. A convolution is an approach to identifying and learning the
features from the image by using an odd-
sized kernel and sliding it over the image.
The convolution operation is performed as follows](https://image.slidesharecdn.com/26194005-250514134919-aacbaa93/85/Intelligent-Data-Analytics-Iot-And-Blockchain-Bashir-Alam-Mansaf-Alam-32-320.jpg)
![Skin Cancer Classification ◾ 9
1.4
Results and Discussion
We notice a higher training accuracy over the ISIC 2020 dataset, because it has more
samples and can learn many more features from the dermoscopic images than other
datasets. We have also implemented regularization techniques, specifically L2 regu-
larization [21], to prevent the model’s overfitting. Figure 1.4 shows the graphical
representation of the LCNet model on the adopted datasets that include ISIC 2016,
ISIC 2017 and ISIC 2020 for training. It displays the performance of the model
Figure 1.4 Train accuracy and loss curves for LCNet on training data. (a) Benign
vs malignant classification for ISIC 2016, (b) benign vs malignant classification for
ISIC 2017, and (c) benign vs malignant classification for ISIC 2020.](https://image.slidesharecdn.com/26194005-250514134919-aacbaa93/85/Intelligent-Data-Analytics-Iot-And-Blockchain-Bashir-Alam-Mansaf-Alam-36-320.jpg)
![16 DOI: 10.1201/9781003371380-2
Chapter 2
Trusted Location
Information Verification
Using Blockchain in
Internet of Vehicles
Ritesh Yaduwanshi and Sushil Kumar
Jawaharlal Nehru University, New Delhi, India
2.1 Introduction
In advanced vehicular adhoc networks (VANETs), a vehicle’s driver can connect to
other vehicles’ drivers, to pedestrians, roadside units and others parts of the urban
infrastructure. VANETs are an intelligent transportation system (ITS) that uses com-
munication to reduce traffic congestion. VANETs can be driven by vehicle-to-road-
side unit (RSU) and/or vehicle-to-vehicle communication [1]. A processing center
(PC) can be an RSU (or a reliable vehicle with a predetermined position). Before
giving directions to the vehicles inside its coverage area, the PC checks the commu-
nication data. Location-based approaches and services are becoming commonplace
in today’s wireless networks, so that location data verification has attracted a lot of
coverage in recent years [1–9]. Generally vehicles get their location from GPS or
GNSS, but the reported location information may be incorrect as the result of either
faulty location data recording/forwarding technology, or deliberate misinformation.
Undesirable network outcomes like insufficient toll payments, traffic congestion or
traffic delays may result if the vehicle’s position information is not validated and
the location inaccuracy is not recognized. In extreme situations, the lack of location](https://image.slidesharecdn.com/26194005-250514134919-aacbaa93/85/Intelligent-Data-Analytics-Iot-And-Blockchain-Bashir-Alam-Mansaf-Alam-43-320.jpg)
![Trusted Location Information Verification ◾ 17
verification could result in disastrous events such as vehicle accidents. These various
location verification systems (LVSs) that have been developed use a range of physi-
cal layer signal parameters to evaluate the vehicle’s reported location information
[2–11]. The constraint of all LVSs is that they typically perform well for the channel
conditions that were considered throughout the design process [2], usually only
working if all of the a priori channel data supplied to them is true. They can also
only properly address the threat-model scenarios for which they were designed [12].
Because of these limitations, their real-world application is limited. The network
can be disrupted simply by vehicles falsifying their location with location-based
access control protocols [10] or geographic routing protocols [11]). A malicious
vehicle can also falsify its location to seriously impair other vehicles [12] and to
enhance its own network capabilities [13]. Accuracy of stated locations in VANETs
is therefore critical and necessitates the use of an LVS. (See [14, 15] for an overview
of IEEE 1609.2 certificate revocation.) This paper is organized as follows: related
work is covered in Section 2.2, Section 2.3 presents the system model, Section 2.4
discusses results, and conclusions are offered in Section 2.5.
2.2 Related Work
Location verification systems are required to overcome the problem of location falsifi-
cation in VANETs. There are two types of existing verification systems: infrastructure-
basedandinfrastructure-less[5,6].Everylocationverificationoperationinthismethod
involves four base stations. Each one counts the time it takes to issue a challenge to
the appropriate node and receive an individual response. However, the large number
of verification requests from automobiles creates a network bottleneck at the base sta-
tion. The cost of deploying and maintaining infrastructure also rises, making VANET
an expensive network An infrastructure-based approach is therefore not suitable for
VANETs. Most infrastructures-less verification systems use various distance measuring
techniques to safely and transparently approve location assertions. For example, in ref.
[7], the solution places verifiers at certain sites, each with its own allowable distance.
The primary constraint of these techniques is the usage of non-RF range technology,
which in turn increases the cost of building these networks. [8] proposes a method for
achieving location verification based solely on logic beacon reception. Autonomous
location verification was implemented in [9] without assuming different trust levels of
nodes. For location verification, [10] uses location-limited channels.
2.3
Trusted Location Information Verification
Using Blockchain
In this section, we present a blockchain-based location verification system model
focusing on location region and location sharping, location sharing and verification
methods, together with the required assumptions.](https://image.slidesharecdn.com/26194005-250514134919-aacbaa93/85/Intelligent-Data-Analytics-Iot-And-Blockchain-Bashir-Alam-Mansaf-Alam-44-320.jpg)
![18 ◾ Intelligent Data Analytics, IoT, and Blockchain
2.3.1 Assumptions
Vehicle-to-everything (V2X) and vehicle-to-vehicle (V2V) communications [11]
are expected, as is the capacity for automobiles to connect to the internet efficiently.
All vehicles are assumed to have the essential equipment, which includes sensors,
GPS and on-board units (OBUs). It is believed that the number of valid roadside
units (RSUs) outnumbers the number of malicious RSUs. We suppose that a valid
RSU constructs a genesis block based on local events to initiate the block chain. We
presume that important event alerts are disseminated within a certain geographic
region of interest (RoI). We presume that the crucial signals are not encrypted, and
that any adjacent vehicles will be able to read them. We assume that 15 messages are
required to validate the occurrence, and that the message is correct.
2.3.2 System Model
Figure 2.1 shows the two stages of the system model according to their functions
of location sharing and location verification. By disclosing the real location coordi-
nates and the real location region during the location sharing step, our suggested
technique ensures that only Owner vehicles can pass location verification. A full
description of the procedure follows.
Table 2.1 Location Verification for Internet of Vehicles
Articles Filtering Cryptography
Infra
structure Verification Detection
Hubauz
et al. [13]
— — Yes Verifiable
multilateration
—
Galle et al.
[18]
— — No Data-processing
model
Errors
explained
Xiao et al.
[19]
— Digital
signature
Yes Signal analysis Statistical
model
Leinmuller
et al. [7]
— — No Trust model
using sensors
—
Yan et al.
[20]
— — No Radar Movement
history
Song et al.
[22]
— Symmetric
keys
No Signal analysis Distance
enlargement
Ren et al.
[21]
Grid
map
— No Filtered data —
Ren et al.
[21]
— — No Directional
antenna
—](https://image.slidesharecdn.com/26194005-250514134919-aacbaa93/85/Intelligent-Data-Analytics-Iot-And-Blockchain-Bashir-Alam-Mansaf-Alam-45-320.jpg)
![Trusted Location Information Verification ◾ 19
2.3.3 Location Sharing
When a Requester asks for the Owner vehicle’s location information, there are two
possibilities: either the Owner vehicle has complete faith in the Requester vehicle
and hence provides precise coordinates (xi, yi), or the Owner vehicle does not com-
pletely trust the Requester vehicle, in which case the Owner vehicle returns the
rectangle location region with location coordinates (xi, yi). The pseudo code for the
location sharing method is shown in Algorithm 1. The Owner vehicle will initially
transmit the public key and session key by asymmetrically encrypting it, computing
Res ← SE(ksession, xi||yi) [23, 24, 25, 26, 27] to encrypt the exact position: Res ←
conRes||ASE(Pubo, ksession). Finally, the session key ksession is retrieved using the
Requester’s private key to decrypt the Owner vehicle’s location coordinates, i.e.,
xi′||yi′ ← SD(ksession, conRes). The privacy-preserving process for location coordinates
is essential when the Owner vehicle does not have total trust in the Requester vehicle
(xi, yi). Through the underchain channel, the Owner vehicle computes fuzRes ←
Enc(ksession, ciphi||borInfo‖nodesx|| nodesy) and provides Res ← fuzRes‖ASE(Pubr,
ksession) to the Requester vehicle. Finally, using the session key ksession produced by
the Requester vehicle’s private key Prir, the Owner vehicle’s location information
is decrypted. The Requester vehicle obtains the privacy-preserving location area
because the border plaintext information {xid1′, xid2′, yid3′, yid4′} is present in borInfo′.
The remainder of fuzRes is used during the Requester vehicle’s location verification.
The location sharing operation for the Owner vehicle and Requester vehicle is now
complete.
Figure 2.1 System model of block chain location verification in internet of
vehicles.](https://image.slidesharecdn.com/26194005-250514134919-aacbaa93/85/Intelligent-Data-Analytics-Iot-And-Blockchain-Bashir-Alam-Mansaf-Alam-46-320.jpg)
![Trusted Location Information Verification ◾ 21
If Hash (xi′||yi′) = LRi·LIi·LHi, it shows that xi′ = xi and yi′ = yi. When the Owner
vehicle has doubts about the Requester, ciphi′||borInfo′||nodes′x||nodes′ySD(ksess
ion, fuzRes) is acquired after the Requester decodes fuzRes; nodes′x and nodes′y are
needed nodes on the root node authentication path for recovering xTree and yTree.
When the Owner has doubts about the Requester, ciphi′||borInfo′||nodes′x||nodes′y
← SD(ksession, fuzRes) is acquired after the Requester decodes fuzRes. The required
nodes on the root node authentication path for recovering xTree and yTree are nodes′x
and nodes′y. First, the Requester checks the received region boundary information
borInfo′ for integrity. If genMT(Hash (borInfoid1′), Hash(borInfoid2′), nodes′x)
= xTreeroot′ and genMT(Hash (borInfoid1′), Hash(borInfoid2′), nodes′y) = yTree-
root′, the regional integrity verification is successful, indicating that the Owner
has returned the proper region information borInfo′ = borInfo. If Hash(ciphi ′) =
LRi·LIi·OpeHashi, (borInfoid1′·ciphx
id1) ≤ (ciphi′·ciphx
i) ≤ (borInfoid2′·ciphx
id2), and
(borInfoid3′·ciphy
id3) ≤ (ciphi ′·Ciphy
i) ≤ (borInfoid4′·ciphy
id4), it shows that xid1 ≤ xi ≤
xid2 and yid3 ≤ yi ≤ yid4; location li is in the region surrounded by {xid1, xid2, yid3, yid4}
that the Requester receives. The region verification is then complete. Finally, the
Owner and Requester’s location verification operation is accomplished. The above
algorithms have computational complexity of O(1) and O(1), respectively. With
increasing N, the compute overhead of Algorithm 1 grows exponentially. Large
plaintext spaces have a high computing overhead, although they can be implemented
offline during the startup process. Furthermore, [28, 29, 30, 31, 32] Algorithm 1
is only used once throughout the entire procedure. The computing overhead of
Algorithm 2 increases as plaintext space grows, with the value of N having no effect.
In the location record phase, it also has a low computing overhead. The two ele-
ments of N and plaintext space have essentially no effect on the processing cost of
Algorithm 1. For Owners, the location sharing phase takes less time. Algorithm 4’s
calculation overhead is approximately linear with N’s size. The plaintext space size
has no bearing on it. Because the location verification phase only involves a hash
operation, the computing overhead is minimal.
ALGORITHM 2 LOCATION VERIFICATION
Input: Location record in the Blockchain (LR); session key Ksession; Location
sharing information (LS)
Output: Boolean variable (LV)
a. Initialize LV ← False
b. If n = 0 then // trust level is maximum
c. xi′||yi′ ← LS;](https://image.slidesharecdn.com/26194005-250514134919-aacbaa93/85/Intelligent-Data-Analytics-Iot-And-Blockchain-Bashir-Alam-Mansaf-Alam-48-320.jpg)
Intelligent Data Analytics Iot And Blockchain Bashir Alam Mansaf Alam
- 1. Intelligent Data Analytics Iot And Blockchain Bashir Alam Mansaf Alam download https://ebookbell.com/product/intelligent-data-analytics-iot-and- blockchain-bashir-alam-mansaf-alam-52388010 Explore and download more ebooks at ebookbell.com
- 2. Here are some recommended products that we believe you will be interested in. You can click the link to download. Intelligent Computing On Iot 20 Big Data Analytics And Block Chain Technology Mohammad S Obaidat Padmalaya Nayak Niranjan K Ray https://ebookbell.com/product/intelligent-computing-on-iot-20-big- data-analytics-and-block-chain-technology-mohammad-s-obaidat- padmalaya-nayak-niranjan-k-ray-56795786 Intelligent Network Design Driven By Big Data Analytics Iot Ai And Cloud Computing Sunil Kumar https://ebookbell.com/product/intelligent-network-design-driven-by- big-data-analytics-iot-ai-and-cloud-computing-sunil-kumar-46225982 Intelligent Network Design Driven By Big Data Analytics Iot Ai And Cloud Computing Kumar https://ebookbell.com/product/intelligent-network-design-driven-by- big-data-analytics-iot-ai-and-cloud-computing-kumar-232127512 Machine Intelligence Big Data Analytics And Iot In Image Processing Ashok Kumar https://ebookbell.com/product/machine-intelligence-big-data-analytics- and-iot-in-image-processing-ashok-kumar-49141322
- 3. Intelligent Data Analytics For Decisionsupport Systems In Hazard Mitigation Theory And Practice Of Hazard Mitigation 1st Ed Ravinesh C Deo https://ebookbell.com/product/intelligent-data-analytics-for- decisionsupport-systems-in-hazard-mitigation-theory-and-practice-of- hazard-mitigation-1st-ed-ravinesh-c-deo-22476414 Intelligent Dataanalytics For Condition Monitoring Smart Grid Applications Hasmat Malik https://ebookbell.com/product/intelligent-dataanalytics-for-condition- monitoring-smart-grid-applications-hasmat-malik-33092976 Intelligent Data Analytics For Power And Energy Systems Hasmat Malik https://ebookbell.com/product/intelligent-data-analytics-for-power- and-energy-systems-hasmat-malik-38429760 Intelligent Data Analytics For Terror Threat Prediction Architectures Methodologies Techniques And Applications Subhendu Kumar Pani https://ebookbell.com/product/intelligent-data-analytics-for-terror- threat-prediction-architectures-methodologies-techniques-and- applications-subhendu-kumar-pani-38117904 Intelligent Dataanalytics For Condition Monitoring Malik Hasmat Fatema https://ebookbell.com/product/intelligent-dataanalytics-for-condition- monitoring-malik-hasmat-fatema-23917788
- 6. This book focuses on data analytics with machine learning using IoT and block- chain technology. Integrating these three fields by examining their interconnections, Intelligent Data Analytics, IoT, and Blockchain examines the opportunities and chal- lenges of developing systems and applications exploiting these technologies. Written primarily for researchers who are working in this multi-disciplinary field, the book also benefits industry experts and technology executives who want to develop their organizations’ decision-making capabilities. Highlights of the book include: ▪ Using image processing with machine learning techniques ▪ A deep learning approach for facial recognition ▪ A scalable system architecture for smart cities based on cognitive IoT ▪ Source authentication of videos shared on social media ▪ Survey of blockchain in healthcare ▪ Accident prediction by vehicle tracking ▪ Big data analytics in disaster management ▪ Applicability, limitations, and opportunities of blockchain technology The book presents novel ideas and insights on different aspects of data analytics, blockchain technology, and IoT. It views these technologies as interdisciplinary fields concerning processes and systems that extract knowledge and insights from data. Focusing on recent advances, the book offers a variety of solutions to real-life challenges with an emphasis on security. Intelligent Data Analytics, IoT, and Blockchain
- 8. Intelligent Data Analytics, IoT, and Blockchain Edited by Bashir Alam Mansaf Alam
- 9. First edition published 2024 by CRC Press 2385 Executive Center Drive, Suite 320, Boca Raton, FL 33431 and by CRC Press 4 Park Square, Milton Park, Abingdon, Oxon, OX14 4RN CRC Press is an imprint of Taylor & Francis Group, LLC © 2024 Taylor & Francis Group, LLC Reasonable efforts have been made to publish reliable data and information, but the author and publisher cannot assume responsibility for the validity of all materials or the consequences of their use. The authors and publishers have attempted to trace the copyright holders of all material reproduced in this publication and apologize to copyright holders if permission to publish in this form has not been obtained. If any copyright material has not been acknowledged please write and let us know so we may rectify in any future reprint. Except as permitted under U.S. Copyright Law, no part of this book may be reprinted, repro- duced, transmitted, or utilized in any form by any electronic, mechanical, or other means, now known or hereafter invented, including photocopying, microfilming, and recording, or in any information storage or retrieval system, without written permission from the publishers. For permission to photocopy or use material electronically from this work, access www.copyright. com or contact the Copyright Clearance Center, Inc. (CCC), 222 Rosewood Drive, Danvers, MA 01923, 978-750-8400. For works that are not available on CCC please contact mpkbookspermis- sions@tandf.co.uk Trademark notice: Product or corporate names may be trademarks or registered trademarks and are used only for identification and explanation without intent to infringe. ISBN: 978-1-032-44278-5 (hbk) ISBN: 978-1-032-44279-2 (pbk) ISBN: 978-1-003-37138-0 (ebk) DOI: 10.1201/9781003371380 Typeset in Garamond by SPi Technologies India Pvt Ltd (Straive)
- 10. v Contents About the Editors...............................................................................................xvii Contributors......................................................................................................xviii 1 Skin Cancer Classification Using Image Processing with Machine Learning Techniques..........................................................1 NIRMALA V., SHASHANK H. S., MANOJ M. M., SATISH ROYAL G., AND PREMALADHA J. 1.1 Introduction.......................................................................................1 1.2 Related Works.....................................................................................2 1.3 Materials and Methods.......................................................................3 1.3.1 Dataset...................................................................................3 1.3.2 Preprocessing Operations.......................................................5 1.3.3 LCNet Architecture................................................................5 1.4 Results and Discussion........................................................................9 1.5 Conclusion.......................................................................................13 References...................................................................................................13 2 Trusted Location Information Verification Using Blockchain in Internet of Vehicles..........................................................16 RITESH YADUWANSHI AND SUSHIL KUMAR 2.1 Introduction.....................................................................................16 2.2 Related Work....................................................................................17 2.3 Trusted Location Information Verification Using Blockchain............17 2.3.1 Assumptions.........................................................................18 2.3.2 System Model......................................................................18 2.3.3 Location Sharing..................................................................19 2.3.4 Location Verification............................................................20 2.4 Results and Simulation.....................................................................22 2.4.1 Location Leakage..................................................................22 2.4.2 Channel Capacity Utilization...............................................22 2.4.3 Message Delivery Success Rate.............................................22 2.4.4 Processing Time...................................................................24
- 11. vi ◾ Contents 2.4.5 Security Attack Resilience.....................................................24 2.5 Conclusion.......................................................................................26 References...................................................................................................26 3 Comparative Analysis of Word-Embedding Techniques Using LSTM Model...............................................................................29 MOHD DANISH AND MOHAMMAD AMJAD 3.1 Introduction.....................................................................................29 3.2 Related Works...................................................................................30 3.3 Methodology....................................................................................31 3.3.1 Dataset.................................................................................31 3.3.2 Word-Embedding Techniques..............................................31 3.3.3 LSTM Deep Learning Classifier...........................................33 3.3.4 Evaluation Metric.................................................................34 3.4 Results and Discussion......................................................................34 3.5 Conclusion and Future Work............................................................35 References...................................................................................................35 4 A Deep Learning Approach for Mask-Based Face Detection..................37 HETA S. DESAI AND ATUL M. GONSAI 4.1 Introduction.....................................................................................37 4.2 Related Work....................................................................................38 4.3 Dataset.............................................................................................40 4.4 Proposed System...............................................................................40 4.4.1 TensorFlow..........................................................................40 4.4.2 Keras....................................................................................40 4.4.3 OpenCV..............................................................................41 4.4.4 Numpy.................................................................................41 4.4.5 Convolution Neural Network (CNN)..................................41 4.5 System Flow Chart............................................................................42 4.6 Evaluating Performance Using Performance Matrix..........................42 4.6.1 Experiments and Result........................................................42 4.7 Conclusion and Future Scope...........................................................46 References...................................................................................................46 5 A Scalable System Architecture for Smart Cities Based on Cognitive IoT.........................................................................................48 NABEELA HASAN AND MANSAF ALAM 5.1 Introduction.....................................................................................48 5.2 Related Work....................................................................................49 5.2.1 IoT Architectural Design......................................................49
- 12. Contents ◾ vii 5.3 Cognitive Computing-based IoT Architecture..................................50 5.3.1 Cognitive Computing-based Smart City Architecture..........51 5.4 Assistive Technologies in Cognitive Computing................................54 5.5 Conclusion.......................................................................................55 References...................................................................................................55 6 Bagging-Based Ensemble Learning for Imbalanced Data Classification Problem...........................................................................57 M. GOVINDARAJAN 6.1 Introduction.....................................................................................57 6.2 Related Work....................................................................................58 6.3 Proposed Methodology.....................................................................59 6.3.1 Pre-processing......................................................................59 6.3.2 Existing Classification Methods............................................59 6.3.3 Homogeneous Ensemble Classifiers......................................59 6.4 Performance Evaluation Measures.....................................................61 6.4.1 Cross-Validation Technique..................................................61 6.4.2 Criteria for Evaluation..........................................................61 6.5 Experimental Results and Discussion................................................61 6.5.1 Vehicle Dataset Description.................................................61 6.5.2 Experiments and Analysis.....................................................62 6.6 Conclusion.......................................................................................63 Acknowledgment........................................................................................63 References...................................................................................................64 7 Design and Implementation of a Network Security Model within a Local Area Network..................................................................65 ADERONKE J. IKUOMOLA, KEHINDE S. OWOPUTI, AND STEPHEN O. JOHNSON-ROKOSU 7.1 Introduction.....................................................................................65 7.1.1 Problem Statement...............................................................66 7.2 Literature Review..............................................................................66 7.3 Design Methodology........................................................................68 7.3.1 Design Consideration...........................................................68 7.3.2 Architecture of a Network Security Model within a LAN.......................................................................69 7.3.3 Software Specification..........................................................70 7.4 Implementation................................................................................70 7.4.1 Network Security Model Implementation Requirements.......................................................................70 7.4.2 The Implemented Local Area Network (LAN) Model and its Configurations..........................................................70
- 13. viii ◾ Contents 7.4.3 Results..................................................................................73 7.5 Conclusion.......................................................................................77 References...................................................................................................77 8 Review of Modern Symmetric and Asymmetric Cryptographic Techniques.............................................................................................79 ANUPAM BHATIA AND NAVEEN NAVEEN 8.1 Introduction.....................................................................................79 8.1.1 Security Services...................................................................80 8.1.2 Cryptography in Data Security.............................................81 8.1.3 Types of Cryptography.........................................................81 8.2 Review of Literature..........................................................................82 8.3 Discussion........................................................................................85 8.4 Conclusion.......................................................................................86 References...................................................................................................87 9 Quantum Computing-Based Image Representation with IBM QISKIT Libraries..........................................................................89 BARKHA SINGH, S. INDU, AND SUDIPTA MAJUMDAR 9.1 Introduction.....................................................................................89 9.2 Objective..........................................................................................90 9.2.1 Main Objective....................................................................90 9.2.2 Algorithm Steps....................................................................90 9.3 Review of Work Implemented..........................................................92 9.3.1 Quantum Circuit of 2n Qubits for a 2 × 2 Image.................94 9.3.2 Tabular Representation of Intensity Values...........................94 9.3.3 Grayscale Image Representation on a Quantum Circuit.................................................................................94 9.4 Advantages........................................................................................97 9.5 Result Analysis..................................................................................97 9.6 Conclusions......................................................................................98 References...................................................................................................99 10 Source Authentication of Videos Shared on Social Media....................102 MOHD SHALIYAR AND KHURRAM MUSTAFA 10.1 Introduction...................................................................................102 10.2 Literature Review............................................................................103 10.3 Proposed Methodology...................................................................105 10.3.1 Watermark Insertion..........................................................105 10.3.2 Watermark Extraction........................................................106 10.4 Experimental Evaluation.................................................................106
- 14. Contents ◾ ix 10.5 Discussion......................................................................................108 10.6 Limitation.......................................................................................112 10.7 Conclusion.....................................................................................112 References.................................................................................................112 11 Task Scheduling Using MOIPSO Algorithm in Cloud Computing......114 RAJESHWARI SISSODIA, MANMOHAN SINGH RAUTHAN, AND VARUN BARTHWAL 11.1 Introduction...................................................................................114 11.2 Related Work..................................................................................116 11.3 Problem Formulation......................................................................117 11.4 System Model.................................................................................118 11.5 Traditional Approach......................................................................118 11.6 Proposed Multi-objective Improved Particle Swarm Optimization..................................................................................120 11.7 Experiment.....................................................................................121 11.7.1 Experimental Set-Up..........................................................121 11.7.2 Experimental Parameters....................................................121 11.7.3 Experiment, Result and Discussion....................................122 11.8 Conclusion and Future Work..........................................................124 References.................................................................................................124 12 Feature Selection-Based Comparative Analysis for Cardiovascular Disease Prediction Using a Machine Learning Model..........................126 SMITA AND ELA KUMAR 12.1 Introduction...................................................................................126 12.2 Related Work..................................................................................127 12.3 Proposed Methodology...................................................................127 12.3.1 Dataset...............................................................................128 12.4 Result Analysis................................................................................130 12.5 Conclusion.....................................................................................133 References.................................................................................................133 13 Use of Cryptography in Networking to Preserve Secure Systems..........135 KAMAL KUMAR, VINOD KUMAR, AND SEEMA 13.1 Introduction...................................................................................135 13.1.1 Characteristics of Cryptography.........................................136 13.1.2 Types of Cryptography.......................................................137 13.1.3 Cryptanalysis......................................................................138 13.2 Cryptographic Primitives................................................................139 13.3 Applications of Cryptography.........................................................140
- 15. x ◾ Contents 13.4 Issues in Network Security..............................................................141 13.5 Issues in Cryptography...................................................................142 13.6 Conclusion and Future Directions..................................................143 References.................................................................................................144 14 Issues and Challenges of Blockchain in Healthcare..............................145 BHAVNA SETHI, HARISH KUMAR, AND SAKSHI KAUSHAL 14.1 Introduction...................................................................................145 14.1.1 Reasons for Adopting Block Chain.....................................145 14.2 Design............................................................................................145 14.2.1 Terms and Definitions........................................................145 14.2.2 Interplanetary File System..................................................146 14.3 Related Work..................................................................................147 14.4 Applications and Challenges of Block Chain in Healthcare.............148 14.4.1 Applications.......................................................................148 14.4.2 Challenges..........................................................................148 14.4.3 Strategies and India-centric Outcomes Targeted towards Block Chain.......................................................................149 14.5 Differences between Current and Proposed Systems........................150 14.5.1 Current System..................................................................150 14.5.2 Proposed System................................................................150 14.5.3 Benefits..............................................................................150 14.5.4 Implementation.................................................................151 14.6 System Architecture........................................................................151 14.7 Conclusion.....................................................................................152 References.................................................................................................153 15 Accident Prediction by Vehicle Tracking..............................................155 GIDDALURI BHANU SEKHAR, JAVVAJI SRINIVASULU, M. BHARGAV CHOWDARY, AND M. SRILATHA 15.1 Introduction...................................................................................155 15.2 Related Work..................................................................................156 15.3 Methodology..................................................................................157 15.3.1 Object Detection and Classification...................................159 15.3.2 Object Tracking..................................................................159 15.3.3 Speed Estimation...............................................................161 15.3.4 Accident Prediction............................................................163 15.4 Results Analysis...............................................................................164 15.5 Performance Analysis......................................................................165 15.6 Conclusion and Future Work..........................................................166 References.................................................................................................167
- 16. Contents ◾ xi 16 Blockchain-Based Cryptographic Model in the Cloud Environment........................................................................................169 PRANAV SHRIVASTAVA, BASHIR ALAM, AND MANSAF ALAM 16.1 Introduction...................................................................................169 16.2 Related Works.................................................................................170 16.3 Proposed Methodology...................................................................172 16.3.1 Protection of Authentication..............................................172 16.3.2 Ownership Protection........................................................172 16.3.3 Identity Mapping Validation..............................................173 16.4 Future Work...................................................................................174 16.5 Conclusions....................................................................................174 References.................................................................................................174 17 Big-Data Analytics in Disaster Management........................................176 PALLAVI AND SANDEEP JOSHI 17.1 Introduction...................................................................................176 17.2 A Disaster-resilience Strategy Based on Big Data.............................177 17.3 Disaster Management.....................................................................178 17.4 Characteristics of Big Data..............................................................180 17.5 Application of Big Data in Disaster Management...........................181 17.6 Comparative Analysis of the Methods Employed............................181 17.7 Conclusion.....................................................................................182 References.................................................................................................182 18 Fuzzy Minimum Spanning Tree Calculation-Based Approach on Acceptability Index Method.................................................................184 PRASANTA KUMAR RAUT, SIVA PRASAD BEHERA, DEBDAS MISHRA, VINOD KUMAR, AND KAMAL LOCHAN MAHANTA 18.1 Introduction...................................................................................184 18.1.1 Literature Review...............................................................185 18.1.2 Motivation and Contribution.............................................185 18.2 Preliminaries...................................................................................186 18.2.1 Triangular Fuzzy Number...................................................186 18.2.2 Trapezoidal Fuzzy Number.................................................186 18.2.3 Yager Index........................................................................186 18.2.4 The π2 Membership Function.............................................187 18.2.5 The Minimum Operation of Two π2-Type Fuzzy Numbers............................................................................187 18.2.6 The Acceptability Index......................................................187 18.2.7 The α-Cut Interval for Fuzzy Number................................188 18.2.8 On α-Cut Interval for Fuzzy Interval..................................189
- 17. xii ◾ Contents 18.2.9 On the Convex Index.........................................................189 18.3 Algorithm for Fuzzy Minimum Spanning Tree................................189 18.3.1 Fuzzy Minimum Spanning Tree Based on the Acceptability Index.............................................................189 18.3.2 Fuzzy Minimum Spanning Tree Algorithm Using Convex Index.....................................................................191 18.3.3 Verification Using Yager’s Index..........................................192 18.3.4 Comparison.......................................................................193 18.4 Conclusion and Future Scope.........................................................193 References.................................................................................................194 19 Encoder/Decoder Transformer-Based Framework to Detect Hate Speech from Tweets.....................................................................195 USMAN AND S. M. K. QUADRI 19.1 Introduction...................................................................................195 19.2 Related Work..................................................................................196 19.3 Preliminaries...................................................................................197 19.3.1 BERT (Bidirectional Encoder Representations from Transformer)......................................................................197 19.3.2 GPT-2 (Generative Pretrained Transformer).......................198 19.4 Framework of the System................................................................198 19.5 Conclusion.....................................................................................203 References.................................................................................................203 20 Understanding Dark Web Protection against Cyber Attacks................208 IRFAN ALAM AND SHAIKH MOHAMMED FAIZAN 20.1 Introduction...................................................................................208 20.2 Elements of the Dark Web..............................................................210 20.2.1 Guard and Middle Relays...................................................212 20.2.2 The Relay is Used to Exit the TOR Circuit.........................212 20.2.3 Bridge................................................................................213 20.3 Criminal Activity............................................................................213 20.3.1 Trafficking..........................................................................213 20.3.2 Information Leakage..........................................................213 20.3.3 Proxying.............................................................................214 20.3.4 Fraud.................................................................................214 20.3.5 Onion Cloning..................................................................214 20.4 Defense Mechanisms and Cyber Attacks.........................................214 20.4.1 Correlation Attacks............................................................214 20.4.2 Congestion Attacks............................................................214 20.4.3 Distributed Denial of Service (DDoS) Attacks...................215 20.4.4 Phishing.............................................................................215
- 18. Contents ◾ xiii 20.4.5 Malware.............................................................................216 20.5 Conclusion.....................................................................................217 References.................................................................................................217 21 Various Elements of Analysis of Authentication Schemes for IoT Devices: A Brief Overview.............................................................219 IRFAN ALAM AND MANOJ KUMAR 21.1 Introduction...................................................................................219 21.2 Motivation......................................................................................221 21.3 Informal Analysis............................................................................222 21.3.1 Adversary Model................................................................222 21.3.2 Taxonomy of Attacks..........................................................224 21.4 Formal Analysis..............................................................................225 21.5 Performance Analysis......................................................................225 21.6 Simulator/Computation Analysis tools...........................................226 21.7 Conclusion and Future Work..........................................................226 Declarations..............................................................................................227 Conflict of Interest....................................................................................227 References.................................................................................................227 22 A Study of Carbon Emissions in the Transport Sector..........................229 AAYESHA ASHRAF AND FARHEEN SIDDIQUI 22.1 Introduction...................................................................................229 22.2 Literature Review............................................................................230 22.3 Data Collection, Analysis and Visualization....................................231 22.4 Technologies for Balancing Emissions.............................................236 22.4.1 Artificial Intelligence (AI)...................................................236 22.4.2 Machine Learning (ML).....................................................237 22.4.3 Internet of Things (IoT).....................................................237 22.4.4 Renewable Energy..............................................................237 22.4.5 Electric Vehicles (EVs)........................................................237 22.4.6 Direct Air Capture (DAC).................................................237 22.4.7 Bioenergy with Carbon Capture and Storage (BECCS)......237 22.5 Conclusion and Future Scope.........................................................238 References.................................................................................................238 23 An Exploration of Blockchain Technology: Applicability, Limitations, and Opportunities...........................................................240 AMARDEEP SAHA AND BAM BAHADUR SINHA 23.1 Introduction...................................................................................240 23.2 Classification of Blockchain............................................................242 23.2.1 Permission-Less Blockchain................................................243
- 19. xiv ◾ Contents 23.2.2 Permissioned Blockchain....................................................243 23.3 Consensus Mechanism....................................................................244 23.3.1 Proof of Work (PoW).........................................................244 23.3.2 Proof of Stake (PoS)...........................................................246 23.3.3 Practical Byzantine Fault Tolerance (PBFT)........................247 23.4 Use Cases of Blockchain Technology...............................................249 23.4.1 Blockchain in the Supply Chain.........................................249 23.4.2 Blockchain for Financial Applications................................249 23.4.3 Blockchain for Non-financial Applications.........................249 23.5 Conclusion and Future Research Areas...........................................249 References.................................................................................................250 24 A Survey of Security Challenges and Existing Prevention Methods in FANET..............................................................................252 JATIN SHARMA AND PAWAN SINGH MEHRA 24.1 Introduction...................................................................................252 24.2 FANET and Communication Protocols..........................................253 24.2.1 Based on Physical Layer......................................................253 24.2.2 Based on MAC Layer.........................................................254 24.2.3 Based on Network Layer/Routing Protocols.......................255 24.3 Security Attacks and Issues..............................................................255 24.3.1 Active Attacks.....................................................................255 24.3.2 Passive Attacks....................................................................255 24.3.3 Other Types of Attack........................................................255 24.4 Literature Review and Related Works..............................................256 24.5 Security Solutions in Tabular Format..............................................258 24.6 Conclusion.....................................................................................260 References.................................................................................................260 25 MENA Sukuk Price Prediction Modeling Using Prophet Algorithm............................................................................................263 TAUFEEQUE AHMAD SIDDIQUI, MOHD RAAGIB SHAKEEL, AND SHAHZAD ALAM 25.1 Introduction...................................................................................263 25.2 Literature Review............................................................................264 25.3 Research Methodology....................................................................268 25.3.1 Prophet Model...................................................................268 25.4 Data Representation.......................................................................269 25.5 Experimental Results and Analyses..................................................270 25.5.1 Evaluation Metrics.............................................................270 25.5.2 Result and Analyses............................................................270 25.6 Conclusion and Implications..........................................................273
- 20. Contents ◾ xv Note���������������������������������������������������������������������������������������������������������273 References.................................................................................................274 26 Cancer Biomarkers Identification from Transcriptomic Data Using Supervised Machine Learning Approaches.................................276 RUBI, FARHAN JALEES AHMAD, BHAVYA ALANKAR, AND HARLEEN KAUR 26.1 Introduction...................................................................................276 26.2 Microarrays in Cancer.....................................................................277 26.3 Supervised Machine Learning in Cancer Biomarkers Detection......278 26.4 Conclusion.....................................................................................279 Acknowledgment......................................................................................283 References.................................................................................................283 27 Development of a Secured and Interoperable Multi-Tenant Software-as-a-Service Electronic Health Record System.......................286 ADERONKE J. IKUOMOLA AND KEHINDE S. OWOPUTI 27.1 Introduction...................................................................................286 27.1.1 Problem Statement.............................................................288 27.2 Literature Review............................................................................288 27.3 Design Methodology......................................................................290 27.3.1 Architecture of a Secured and Interoperable Multi-tenant SaaS Electronic Health Record System..........290 27.3.2 Components of the Architectural Design...........................290 27.3.3 Flowchart...........................................................................292 27.4 Implementation..............................................................................292 27.4.1 The Security Framework.....................................................298 27.5 Conclusion.....................................................................................299 References.................................................................................................300 28 Investigating Classification with Quantum Computing.......................302 MUHAMMAD HAMID, BASHIR ALAM, OM PAL, AND SHAMIMUL QAMAR 28.1 Introduction...................................................................................302 28.2 Quantum Computation Background..............................................303 28.2.1 Circuits and Measurements................................................306 28.3 Quantum Machine Learning..........................................................306 28.3.1 Quantum Encoding...........................................................307 28.4 Literature Review............................................................................308 28.5 Quantum Machine Learning Algorithms........................................310 28.6 Challenges and Future Scope..........................................................311
- 21. xvi ◾ Contents 28.7 Conclusion.....................................................................................311 References.................................................................................................312 29 A Comprehensive Analysis of Techniques Offering Dynamic Group Management in a Cloud Computing Environment...................315 PRANAV SHRIVASTAVA, BASHIR ALAM, AND MANSAF ALAM 29.1 Introduction...................................................................................315 29.2 Existing Solutions Based on Encryption Mechanisms.....................316 29.3 Kerberos-Based Solutions................................................................319 29.4 Access Control-Based Solutions......................................................320 29.5 Conclusion.....................................................................................321 References.................................................................................................322 30 Improved YOLOv5 with Attention Mechanism for Real-Time Weed Detection in the Paddy Field: A Deep Learning Approach.........326 BHUVANESWARI SWAMINATHAN, PRABU SELVAM, JOSEPH ABRAHAM SUNDAR K., AND SUBRAMANIYASWAMY VAIRAVASUNDARAM 30.1 Introduction...................................................................................326 30.2 Related Works.................................................................................328 30.3 Proposed System.............................................................................328 30.3.1 Improved YOLOv5 Algorithm...........................................328 30.3.2 Attention Mechanism.........................................................331 30.3.3 CBAM...............................................................................332 30.3.4 ECA-Net............................................................................333 30.4 Experiments....................................................................................333 30.4.1 Implementation Details......................................................333 30.4.2 Evaluation Metrics.............................................................333 30.4.3 Training.............................................................................334 30.4.4 Ablation Studies.................................................................334 30.5 Performance Analysis......................................................................338 30.5.1 Comparison with State-of-the-Art Approaches...................338 30.6 Conclusion.....................................................................................340 References.................................................................................................340 Index������������������������������������������������������������������������������������������������������������342
- 22. xvii About the Editors Bashir Alam, PhD, is a professor at Jamia Millia Islamia, New Delhi, India, where he heads the Department of Computer Engineering. He has 22 years of teaching and research experience. His areas of research include big-data analytics, artificial intelli- gence, parallel and distributed systems, cloud computing, machine learning, GPU com- puting, blockchain, and information security. Mansaf Alam, PhD, is a professor in the Department of Computer Science, Faculty of Natural Sciences, Jamia Millia Islamia. A Young Faculty Research Fellow and the editor- in-chief of the Journal of Applied Information Science, he pursues research in artificial intelligence, big-data analytics, machine learning, deep learning, cloud computing, and data mining.
- 23. xviii Contributors Farhan Jalees Ahmad School of Interdisciplinary Sciences and Technology, Jamia Hamdard New Delhi, India Bashir Alam Department of Computer Engineering, Jamia Millia Islamia University New Delhi, India Irfan Alam Department of Computer Science and Engineering, Delhi Technological University New Delhi, India Mansaf Alam Department of Computer Sciences, Jamia Millia Islamia University New Delhi, India Shahzad Alam Department of Computer Engineering, Faculty of Engineering and Technology, Jamia Millia Islamia University New Delhi, India Bhavya Alankar Department of Computer Science and Engineering, School of Engineering Sciences and Technology, Jamia Hamdard New Delhi, India Mohammad Amjad Jamia Millia Islamia University New Delhi, India Aayesha Ashraf Department of Computer Science and Engineering, Jamia Hamdard (Deemed to be University) New Delhi, India Varun Barthwal H.N.B. Garhwal University Srinagar, India Siva Prasad Behera Department of Mathematics, C.V. Raman Global University Bhubaneswar, India Anupam Bhatia CRSU Jind, India
- 24. Contributors ◾ xix M. Bhargav Chowdary Jawaharlal Nehru Technological University Kakinada, India Mohd Danish Jamia Millia Islamia University New Delhi, India Heta S. Desai Saurashtra University Rajkot, India Shaikh Mohammed Faizan Department of Computer Engineering, Jamia Millia Islamia University New Delhi, India Atul M. Gonsai Saurashtra University Rajkot, India M. Govindarajan Department of Computer Science and Engineering, Annamalai University Annamalai Nagar, India Muhammad Hamid Department of Computer Engineering, Jamia Millia Islamia University New Delhi, India Nabeela Hasan Jamia Millia Islamia University New Delhi, India Aderonke J. Ikuomola Department of Computer Science, Olusegun Agagu University of Science and Technology Okitipupa, Nigeria S. Indu Delhi Technological University (AICTE) Delhi, India Stephen O. Johnson-Rokosu Olusegun Agagu University of Science and Technology Okitipupa, Nigeria Dr. Sandeep Joshi Department of Computer Science and Engineering, Manipal University Jaipur Jaipur, India Joseph Abraham Sundar K. School of Computing, SASTRA Deemed University Thanjavur, India Harleen Kaur Department of Computer Science and Engineering, School of Engineering Sciences and Technology, Jamia Hamdard New Delhi, India Sakshi Kaushal Department of Computer Science and Engineering, UIET, Panjab University Chandigarh Chandigarh, India Sudipta Majumdar Delhi Technological University (AICTE) India Delhi, India Ela Kumar Indira Gandhi Delhi Technical University for Women New Delhi, India
- 25. xx ◾ Contributors Harish Kumar Department of Computer Science and Engineering, UIET, Panjab University Chandigarh Chandigarh, India Kamal Kumar Department of Mathematics, Baba Mastnath University Rohtak, India Sushil Kumar School of Computer Systems Sciences, Jawaharlal Nehru University New Delhi, India Vinod Kumar Department of Mathematics, PGDAV Collage, University of Delhi New Delhi, India Manoj M. M. School of Computing, SASTRA Deemed to be University India Manoj Kumar Department of Computer Science and Engineering, Delhi Technological University New Delhi, India School of Computing, SASTRA Deemed to be University Thanjavur, India Kamal Lochan Mahanta Department of Mathematics, C.V. Raman Global University Bhubaneswar, India Pawan Singh Mehra Department of Computer Science and Engineering, Delhi Technological University New Delhi, India Debdas Mishra Department of Mathematics, C.V. Raman Global University Bhubaneswar, India Barkha Singh ECE Dept (Of AICTE), Delhi Technological University (AICTE Delhi, India Khurram Mustafa Department of Computer Science, Jamia Millia Islamia University New Delhi, India Naveen Naveen CRSU Jind, India Kehinde S. Owoputi Department of Computer Science, Olusegun Agagu University of Science and Technology Okitipupa, Nigeria Om Pal MeitY, Government of India New Delhi, India Pallavi Department of Computer Science and Engineering, Manipal University Jaipur Jaipur, India Premaladha J. School of Computing, SASTRA Deemed to be University Thanjavur, India Shamimul Qamar Department of Computer Science Engineering, King Khalid University Abha, Kingdom of Saudi Arabia
- 26. Contributors ◾ xxi S. M. K. Quadri Jamia Millia Islamia University New Delhi, India Prasanta Kumar Raut Department of Mathematics, C.V. Raman Global University Bhubaneswar, India Manmohan Singh Rauthan H.N.B. Garhwal University Srinagar, India Rubi School of Interdisciplinary Sciences and Technology, Jamia Hamdard New Delhi, India Seema Department of Mathematics, Baba Mastnath University Rohtak, India Amardeep Saha Computer Science and Engineering, Indian Institute of Information Technology Ranchi Ranchi, India Department of Mathematics, Baba Mastnath University Rohtak, India Shashank H. S. School of Computing, SASTRA Deemed to be University Thanjavur, India Satish Royal G. School of Computing, SASTRA Deemed to be University Thanjavur, India Giddaluri Bhanu Sekhar Jawaharlal Nehru Technological University Kakinada Kakinada, India Prabu Selvam School of Computing, SASTRA Deemed University Thanjavur, India Bhavna Sethi UIET, Punjab University Chandigarh Chandigarh, India Mohd Raagib Shakeel Department of Management Studies, Jamia Millia Islamia University New Delhi, India Mohd Shaliyar Department of Computer Science, Jamia Millia Islamia University Department of Computer Science and Engineering, Delhi Technological University New Delhi, India Jatin Sharma Department of Computer Science and Engineering, Delhi Technological University New Delhi, India Pranav Shrivastava Department of Computer Engineering, JMI New Delhi, India Farheen Siddiqui Department of Computer Science and Engineering Jamia Hamdard (Deemed to be University) New Delhi, India
- 27. xxii ◾ Contributors Taufeeque Ahmad Siddiqui Department of Management Studies, Jamia Millia Islamia University New Delhi, India Bam Bahadur Sinha Computer Science and Engineering, Indian Institute of Information Technology Ranchi Ranchi, India Rajeshwari Sissodia H.N.B. Garhwal University Srinagar, India Smita CSE, Indira Gandhi Delhi Technical University for Women New Delhi, India M. Srilatha Jawaharlal Nehru Technological University Kakinada Kakinada, India Javvaji Srinivasulu Jawaharlal Nehru Technological University Kakinada Kakinada, India Bhuvaneswari Swaminathan School of Computing, SASTRA Deemed University Thanjavur, India Usman Jamia Millia Islamia University New Delhi, India Nirmala V. School of Computing, SASTRA Deemed to be University Thanjavur, India Subramaniyaswamy Vairavasundaram School of Computing, SASTRA Deemed University Thanjavur, India Ritesh Yaduwanshi School of Computer Systems Sciences, Jawaharlal Nehru University New Delhi, India
- 28. 1 DOI: 10.1201/9781003371380-1 Chapter 1 Skin Cancer Classification Using Image Processing with Machine Learning Techniques NirmalaV., Shashank H. S., Manoj M. M., Satish Royal G., and Premaladha J. School of Computing, SASTRA Deemed to be University, Thanjavur, India 1.1 Introduction Image classification modalities play a significant role in the health sector. Early diag- nosis of fatal diseases using various imaging techniques [1] has positively impacted people’s lives. Our work describes classification of skin cancer images using deep learning techniques [2]. Skin cancer attacks surrounding cells, resulting in the devel- opment of a mole on the external layer of the skin that can be categorized as malig- nant or benign. Many solutions using neural network architectures for diagnosis of the early stages of skin cancer have been proposed. The classification [3] metrics used include support vector machine (SVM), relevant vector machine (RVM), and neural network architectures. These machine learning algorithms pose several con- straints for input data distribution, such as noise- free or high- contrast images, but these constraints do not apply to the skin cancer classification problem. Instead, it
- 29. 2 ◾ Intelligent Data Analytics, IoT, and Blockchain is colour, texture, and structural features that play an essential role in skin cancer classification. Traditional parametric approaches cannot be used for skin cancer clas- sification problems since skin lesions have different patterns. Hence deep learning techniques are used. The automatic classification process [4] includes preprocessing, feature extrac- tion, segmentation, and classification, resulting in a handcrafted feature set. However, since lesions have visual resemblance and are highly correlated due to their colour, texture, and shape, handcrafted feature extraction is not appropriate for skin cancer classification. The deep learning approach is therefore preferred. We can feed the images directly to the model, removing the need for any preprocessing [5] to be implemented before passing the image to the model. Neural network models are very effective in extracting specific features from the image. Even though deep learning models are efficient for classifying skin cancer, the various elements present in skin lesion images make identifying skin cancer [6] challenging for the following reasons: ▪ The ISIC 2016 and ISIC 2017 skin cancer datasets are highly imbalanced, with many benign samples. ▪ Many skin lesion images are highly similar, and classifying [7] them into benign and malignant images is challenging. The novel modified LCNet model is designed for model training for boosted clas- sification results even for the less accurate lesions of human skin. An optimization algorithm is improved with the repeated blocks of batch normalization. The remainder of this chapter is organized as follows. Section 1.2 describes sig- nificant earlier work in this area. Section 1.3 explains our research in detail, the data- sets involved and the model architecture. Sections 1.4 and 1.5 present our results with their comparative analysis and conclusion. 1.2 Related Works Skin cancer is a deadly disease that can affect nearby cells of the body. Early detection and diagnosis is important [8]. Initially, handcrafted feature- based approaches were used on dermoscopic images. However, since there is a high correlation between skin texture and colour in skin images, such approaches are not regarded as suitable for skin classification problems [9]. As preprocessing operations are unnecessary, deep convolutional neural networks (DCNNs) have proved helpful. The first time DCNN was applied to skin cancer images [10] used 129,450 skin disease images to classify 2032 diseases. The researcher designed a deep learning framework with two fully convolutional residual networks, one to produce the segmentation result and the other to produce a coarse classification result. A lesion index calculation [11] unit was introduced to produce a heat map and refine the coarse classification results. Iqbal et al. [12] describe the contribution of each pixel towards the classifica- tion of another model, a convolution model consisting of multiple layers used for
- 30. Skin Cancer Classification ◾ 3 multi- class classification. It had 68 convolutional layers, passing the features from top to bottom. Zhang et al. proposed an attention residual neural network [13] that consisted of multiple ARL blocks, which was further followed by global average pooling and classification layers. To improve classification efficiency, ensembles of CNNs, consisting of outputs from the layers of four different models, Google Net, AlexNet, VGG and ResNet, were created by Barata et al. [14]. Another approach using multiple imaging modali- ties was also proposed to increase the modularity of a self- supervised topology clus- tering network that could classify the unlabeled data without needing class- based information. The model learnt features at different levels of variations, such as the illumination, the background and the point of view. Some models applied transfer learning [15] using pre- trained models for dermo- scopic classification. Nevertheless, a small dataset with high accuracy does not fit all scenarios, especially with medical images, since each piece of information is highly sensitive in the diagnosis. Hyperparameter tuning was performed to achieve better results. Gessert et al. developed an ensemble model from Efficient Nets, SENet, and ResNet WSL which was used to perform a multi- class classification task [16] on the ISIC 2019 dataset. A cropping strategy was implemented to deal with different input resolutions. With the earlier reports for melanoma classification, many researchers carried out different trials. They achieved reasonable accuracy, and some of the results are inspiring. However, the status of early diagnosis of melanoma skin cancer is not generalized and, as we have seen in the Introduction, there are two major problems. 1.3 Materials and Methods 1.3.1 Dataset The skin cancer images were obtained from the ISIC 2016 [17], ISIC 2017 [18] and ISIC 2020 [19] challenges. Since the skin samples were highly imbalanced, they were augmented by datasets from the ISIC archive. The ISIC challenge provides two datasets – training and testing. We divided our model learning and estimation process into three parts: training, validation and testing. We chose 20% for the vali- dation process. Since the given skin cancer samples were highly imbalanced, data augmentation was carried out on classes with fewer samples to prevent the deteriora- tion of the model learning process. Data were classified into two classes: MEL (malignant) and BEN (benign). Other lesion types, such as seborrheic keratosis and nevus, are also considered benign. The total number of training samples and validation samples is shown in Table 1.1. The data distribution over ISIC 2016, 2017 and 2020 datasets is shown in Figure 1.1.
- 31. 4 ◾ Intelligent Data Analytics, IoT, and Blockchain Table 1.1 Size of the Datasets Dataset Total Number of Samples Training Samples (80%) of Total Samples Validation (20%) of Total Samples Samples Under Test Data ISIC 2016 900 720 180 379 ISIC 2017 1620 1296 324 600 ISIC 2020 33126 26500 6626 439 Figure 1.1 Data distribution over ISIC 2016, 2017 and 2020 datasets; cross- hatched – benign samples, shaded – malignant samples.
- 32. Skin Cancer Classification ◾ 5 1) The proposed model uses a DCNN for classifying images into benign and malignant. The model consists of multiple blocks bonded together to facili- tate the processing of many features in the convolutional neural network architecture. 2) Each block consists of varying parameters having different values. Parameters include stride, number of kernels, and kernel size. 3) The model consists of 11 blocks, each having its sequence of operations per- formed over the image. 1.3.2 Preprocessing Operations For the given data sets, preprocessing was carried out to make them suitable for passing through the model. The images were normalized to make the computations effective. The normalization of images was carried out using pixel normalization. The pixel values were scaled to 0–1. Normalization is essential because it ensures that each input parameter has a similar data distribution. Furthermore, data augmentation – rotation, shifting, flipping, and scaling – was carried out since the image samples were highly imbalanced [20]. A random rota- tion of 0° to 90° was applied to the image. The image was shifted by 10% of the entire width and height. Horizontal and vertical flipping was carried out on all the images. These augmentation operations [21] were applied only to the training and validation sets. Sample augmentation operations are shown in Figure 1.2. Since the model accepts a (128, 128, 3) image, all the images in the dataset were resized [22] to (128, 128, 3). For the model, we stored all the images in HDF5 format and organized them into folders depending on the type and category of the image [23]. 1.3.3 LCNet Architecture The proposed DCNN model, Lesion Classification Network (LCNet), is formu- lated using 11 blocks, as shown in Figure 1.3. The model also has the following specification: ▪ Block 4 and 5 – repeated twice ▪ Block 7 and 8 – repeated 4 times ▪ Blocks 10 and 1 – repeated twice The network accepts a (128 × 128 × 3) image as an input, after which a convolu- tion operation is performed over the image using a (3 × 3) kernel having a stride of 2 to learn 8 features. A convolution is an approach to identifying and learning the features from the image by using an odd- sized kernel and sliding it over the image. The convolution operation is performed as follows
- 33. 6 ◾ Intelligent Data Analytics, IoT, and Blockchain Conv u v h i j F u i v j i k k , , . , (1.1) Here Conv(u, v) is the output of the convolution operation on the image using a kernel whose pixel positions are identified by using (i, j). Moreover, “k” determines the maximum size of the kernel in positive and negative axes. The h(i, j) is the ker- nel, and F(u, v) represents the pixel locations of the original image. The output of a convolution is a feature map that is reduced by passing to a max- pooling layer. The max- pooling layer takes a pool size of (2, 2) and identifies the maximum pixel value in each pool. Further, each block has three essential layers: 1) Convolution 2) Batch normalization 3) LeakyReLU (leaky rectified linear unit) Figure 1.2 Augmentation operation on skin images (a) width shift, (b) height shift, (c) flipping, and (d) zooming.
- 34. Skin Cancer Classification ◾ 7 The input features to the subsequent blocks are normalized using batch normal- ization, a technique used for training deep neural network architecture. Here the inputs to a layer from the previous one are standardized for each mini- batch. These servers increase the learning process and speed up the training. The activation func- tion used here is ‘LeakyReLU’ – leaky rectified linear unit as given by f x s x x x x , , 0 0 (1.2) The LeakyReLU overcame the ‘dying ReLU’ problem when x is less than zero. This blocks the process of learning in the ReLU. LeakyReLU speeds up the training process, as having a mean activation close to zero makes the training faster. Moreover, the LeakyReLU does not have a zero slope. The main advantage of LeakyReLU can be seen when during backpropagation, the weights are to be updated. Figure 1.3 The model for classification of skin cancer – LCNet.
- 35. 8 ◾ Intelligent Data Analytics, IoT, and Blockchain In ReLU, some dead neurons may never activate again, so training them wastes time. Our model uses a scaling factor ′s′ value of 0.3. Figure 1.3 shows the LCNet architecture. Block 1 comprises two convolution layers, two batch- normalization and two LeakyReLU layers. The convolution layers consist of the following: ▪ First convolution layer – 16 kernels of size (1,1) and stride 1 ▪ Second convolution layer – 32 kernels of size (1,1) and stride 1. The stride determines the number of pixels by which the kernel moves. A batch normalizer and LeakyReLU succeed in each convolution layer. Block 2 consists of the following: ▪ Single convolution layer – 32 kernels, size (3,3), and stride 1 The result of the previous convolution is a feature map succeeded by LeakyReLU and batch normalizer. The output features of max- pooling, Block1 and Block2 are combined and are then passed to Block 3. Other blocks of the DCNN are con- structed in a similar manner, having a different number of filters and kernel sizes. Towards the end of the neural network, a global average pooling layer is used, fol- lowed by a 2x fully connection layer. The model uses the stochastic gradient descent (SGD) algorithm for the opti- mization of the model’s parameters. The SGD algorithm is used to minimize the loss function, and to reach a global minimum such that the output is closest to the required value. The model has a learning rate of 0.0005. The learning rate is required for the reduction of the loss of the model in SGD, which is achieved by modifying the model weights. A very high learning rate may increase loss, while a low learning rate may require more iterations. Furthermore, the model uses a categorical- cross entropy loss function. The proposed model uses an SGD optimizer to modify and update the neural network’s weights during backpropagation. It is essential to minimize the error gradi- ent, find the model parameters that produce an outcome and be closely related to the actual output. Table 1.2 shows all the hyperparameters used in the model training. Table 1.2 Hyperparameters of the Model Mini- Batch Size Data Augmen tation Regular ization Value Optimi zation Algorithm Learning Rate Momentum Activation Function Epochs 32 Flipping, Rotation, Shifting, Scaling 0.0005 SGD 0.0005 0.99 LeakyReLU 50
- 36. Skin Cancer Classification ◾ 9 1.4 Results and Discussion We notice a higher training accuracy over the ISIC 2020 dataset, because it has more samples and can learn many more features from the dermoscopic images than other datasets. We have also implemented regularization techniques, specifically L2 regu- larization [21], to prevent the model’s overfitting. Figure 1.4 shows the graphical representation of the LCNet model on the adopted datasets that include ISIC 2016, ISIC 2017 and ISIC 2020 for training. It displays the performance of the model Figure 1.4 Train accuracy and loss curves for LCNet on training data. (a) Benign vs malignant classification for ISIC 2016, (b) benign vs malignant classification for ISIC 2017, and (c) benign vs malignant classification for ISIC 2020.
- 37. 10 ◾ Intelligent Data Analytics, IoT, and Blockchain on the training data. We observe that, as the training accuracy increases, the loss reduces. The following models were trained over 50 epochs. Early stopping criteria were implemented to assess the model’s result in case the training accuracy did not improve in successive epochs. Figure 1.5 shows the graphical representation of the LCNet architecture on the validation data. It displays the network’s performance on the validation set in terms of accuracy and loss. We observe that the validation accuracy is a constant curve in most Figure 1.5 Validation accuracy and loss curves for the LCNet on the validation. (a) Benign vs malignant classification for ISIC 2016, (b) benign vs malignant classification for ISIC 2017, and (c) benign vs malignant classification for ISIC 2020.
- 38. Skin Cancer Classification ◾ 11 cases. This is because the data distribution of the adopted datasets was imbalanced. However, we achieve an optimal accuracy nearing 0.8. Also, we can confirm that the model can learn, as the validation loss is minimal towards the end of the epochs. In this work, we have used major four classification metrics: 1. Accuracy (ACC) 2. Precision (PRE) 3. Recall (REC) 4. F1-Score (F1) The mathematical formulae defining the above metrics are as follows: ACC TP TN TP TN FP FN (1.3) PRE TP TP FP (1.4) REC TP TP FN (1.5) F PRE REC PRE REC 1 2 (1.6) We observe that the first two confusion matrices are highly biased in identifying benign skin cancers since the data distribution used to train the model was imbal- anced. However, in the last confusion matrix, we observe an optimal performance where 49 skin samples were correctly classified as malignant, and 160 were classified as benign. Figure 1.6 displays a confusion matrix that consists of four boxes. This evalua- tion is used to measure the performance of our classification, which shows the true and false results. ▪ The first box (top left corner) exhibits the number of skin lesion images classi- fied correctly as benign, which depicts the valid positive rate. ▪ The second box (top right corner) represents the number of input lesions mis- predicted as malignant. It represents the false negative rate. ▪ The third box (bottom left corner) depicts the number of lesion images incor- rectly predicted as benign. It represents the false positive rate. ▪ The fourth box (bottom right corner) represents total image lesions correctly classified as malignant and eventually represents the valid negative rate.
- 39. 12 ◾ Intelligent Data Analytics, IoT, and Blockchain Since in the ISIC 2016 and ISIC 2017 datasets, the total number of benign samples was far greater than the number of malignant samples, the model could learn the benign features accurately. Hence, the confusion matrix for ISIC 2016 and ISIC 2017 is biased towards benign samples. Table 1.3 shows the performance metrics of classification using the LCNet architecture. Table 1.4 compares our proposed model with other state-of-the-art models. The experimental results in Table 1.4 show that the model performs relatively well in classifying benign and malignant skin cancers. We observe a better accuracy in ISIC 2017 dataset compared to the remaining models. However, with oversam- pling, much better accuracy can be achieved as the model can learn more features regarding the malignant samples. Our model requires many samples to distinguish between malignant and benign. An advantage that our model has over other models is its significantly smaller number of parameters, which make it a low- weight model. Figure 1.6 Confusion matrix of LCNet on test data (a) benign vs malignant classification for ISIC 2016, (b) benign vs malignant classification for ISIC 2017, and (c) benign vs malignant classification for ISIC 2020.
- 40. Skin Cancer Classification ◾ 13 1.5 Conclusion The proposed LCNet architecture has been trained over several skin lesion samples to learn features and help detect melanoma skin cancer. The model has been trained and tested over three datasets: ISIC 2016, 2017 and 2020. The model has an accu- racy of 92.130%, 91.5% and 91.43% on the given datasets. The model achieves good accuracy in terms of the classification of benign and malignant skin cancers. Since many benign samples greatly influenced the model’s training over the ISIC 2016 and 2017 datasets, the model was seen to be biased towards benign sam- ples. The model’s learning rate can be further increased by adding more malignant samples to the dataset. This random oversampling can further improve the model’s prediction. References 1. American Cancer Society. Key Statistics for Melanoma Skin Cancer. 2021. Available online: https://www.cancer.org/cancer/%20melanoma-skin-cancer/about/key-statistics.html (accessed on 15 December 2020). Table 1.3 Performance Evaluation of LCNet on the Adopted Dataset Without Oversampling ISIC 2020 ISIC 2017 ISIC 2016 ACC% 91.43 91.5 92.031 PRE% 86.8 80.5 74.52 REC% 43.0 98.0 95.43 F1% 57.421 89.19 83.69 Table 1.4 Experimental Results of Our ProposedWork on theAdopted Dataset Approach ISIC 2020 ACC PRE REC ISIC 2017 ACC PRE REC ISIC 2016 ACC PRE REC ResNet18 0.908 0.898 0.888 0.750 0.640 0.571 0.809 0.789 0.809 Inceptionv3 0.486 0.297 0.492 0.774 0.691 0.612 0.799 0.809 0.811 Alex Net 0.754 0.691 0.685 0.740 0.670 0.660 0.654 0.595 0.64 Proposed Model (LCNet) 0.91 0.86 0.43 0.915 0.805 1.00 0.92 0.74 0.95
- 41. 14 ◾ Intelligent Data Analytics, IoT, and Blockchain 2. Rahi, Md. Muzahidul Islam; Khan, Farhan Tanvir; Mahtab, Mohammad Tanvir; Ullah, A. K. M. Amanat; Alam, Md. Golam Rabiul; Alam, Md. Ashraful. Detection of skin cancer using deep neural networks. 2019: https://ieeexplore.ieee.org/abstract/document/9162400/ authors#authors 3. Jinnai, S.; Yamazaki, N.; Hirano, Y.; Sugawara, Y.; Ohe, Y.; Hamamoto, R. The develop- ment of a skin cancer classification system for pigmented skin lesions using deep learning. Biomolecules. 2020, 10, 1123: http://dx.doi.org/10.3390/biom10081123 4. Liu, L.; Mou, L.; Zhu, X.X.; Mandal, M. Automatic skin lesion classification based on mid- level feature learning. Comput. Med. Imaging Graph. 2020, 84, 101765: http://dx.doi. org/10.1016/j.compmedimag.2020.101765 5. Kwasigroch, A.; Grochowski, M.; Mikołajczyk, A. Neural architecture search for skin lesion classification. IEEE Access 2020, 8, 9061–9071: http://dx.doi.org/10.1109/ ACCESS.2020.2964424 6. Tharwat, A. Classification assessment methods. Appl. Comput. Inform. 2020, 17, 168–192: http://dx.doi.org/10.1016/j.aci.2018.08.003 7. Tang, P.; Liang, Q.; Yan, X.; Xiang, S.; Zhang, D. GP- CNN- DTEL: Global- part CNN model with data-transformed ensemble learning for skin lesion classification. IEEE J. Biomed. Health Inform. 2020, 24, 2870–2882: http://dx.doi.org/10.1109/JBHI.2020.2977013, http://www.ncbi.nlm.nih.gov/pubmed/32142460 8. International Agency for Research on Cancer. Cancer – World Health Organization. 2020. Available online: https://www.who.int/cancer/PRGlobocanFinal.pdf (accessed 15 December 2020). 9. Thomsen, K.; Iversen, L.; Titlestad, T.L.; Winther, O. Systematic review of machine learn- ing for diagnosis and prognosis in dermatology. J. Dermatol. Treat. 2020, 31, 496–510: http://dx.doi.org/10.1080/09546634.2019.1682500 10. Al- Masni, M. A.; Kim, D. H.; Kim, T. S. Multiple skin lesions diagnostics via integrated deep convolutional networks for segmentation and classification. Comput. Methods Programs Biomed., 2020, 190, 105351. 11. Li, Y.; Shen, L. Skin lesion analysis towards melanoma detection using deep learning network. Sensors (Basel). 2018 February. 11, 18(2), 556. 10.3390/s18020556. PMID: 29439500; PMCID: PMC5855504. 12. Iqbal, I.; Younus, M.; Walayat, K.; Kakar, M.U.; Ma, J. Automated multi- class classifica- tion of skin lesions through deep convolutional neural network with dermoscopic images. Comput. Med. Imaging Graph. 2021, 88, 101843. 13. Zhang, J.; Xie, Y.; Xia, Y.; Shen, C. Attention residual learning for skin lesion classification. IEEE Trans. Med. Imaging 2019, 38, 2092–2103. 14. Subramanian, R Raja; Achuth, Dintakurthi; Kumar, P Shiridi; Kumar Reddy, Kovvuru Naveen; Amara, Srikar; Chowdary, Adusumalli Suchan. Skin cancer classification using convolutional neural networks. 2020: https://ieeexplore.ieee.org/document/ 9377155/authors 15. Ashraf, R.; Afzal, S.; Rehman, A.U.; Gul, S.; Baber, J.; Bakhtyar, M., et al. Region- of- interest based transfer learning assisted framework for skin cancer detection. IEEE Access, 2020, 8, 147858–147871. 16. Rotemberg, V.; Kurtansky, N.; Betz- Stablein, B.; Caffery, L.; Chousakos, E.; Codella, N.; Combalia, M.; Dusza, S.; Guitera, P.; Gutman, D.; et al. A patient- centric dataset of images and metadata for identifying melanomas using clinical context. Sci. Data. 2021, 8, 34: http://dx.doi.org/10.1038/s41597-021-00815-z
- 42. Skin Cancer Classification ◾ 15 17. Gutman, David; Codella, Noel C. F.; Celebi, Emre; Helba, Brian; Marchetti, Michael; Mishra, Nabin; Halpern, Allan. Skin Lesion Analysis toward Melanoma Detection: A Challenge at the International Symposium on Biomedical Imaging (ISBI) 2016, hosted by the International Skin Imaging Collaboration (ISIC). eprint arXiv:1605.01397. 2016. 18. Codella, N.; Gutman, D.; Celebi, M.E.; Helba, B.; Marchetti, M.A.; Dusza, S.; Kalloo, A.; Liopyris, K.; Mishra, N.; Kittler, H.; Halpern, A. Skin Lesion Analysis Toward Melanoma Detection: A Challenge at the 2017 International Symposium on Biomedical Imaging (ISBI), Hosted by the International Skin Imaging Collaboration (ISIC). arXiv: 1710.05006, 2018. 19. International Skin Imaging Collaboration. SIIM- ISIC 2020 challenge dataset. International Skin Imaging Collaboration, 2020, https://doi.org/10.34970/2020-ds01 20. Premaladha, J.; Surendra Reddy, M.; Hemanth Kumar Reddy, T.; Sri Sai Charan, Y.; Nirmala, V. Recognition of Facial Expression Using Haar Cascade Classifier and Deep Learning. In: Ranganathan, G., Fernando, X., Shi, F. (eds) Inventive Communication and Computational Technologies. Lecture Notes in Networks and Systems, vol. 311. Springer, Singapore, 2022. https://doi.org/10.1007/978-981-16-5529-6_27 21. Medhat, Sara; Abdel- Galil, Hala; Aboutabl, Amal Elsayed; Saleh, Hassan. Skin cancer diag- nosis using convolutional neural networks for smartphone images: A comparative study. J. Radiat. Res. Appl. Sci. 2022, 15(1), 262–267, ISSN 1687-8507, https://doi.org/10.1016/j. jrras.2022.03.008 22. Jayaraman, P., Veeramani, N., Krishankumar, R., Ravichandran, K. S., Cavallaro, F., Rani, P., Mardani, A. Wavelet- based classification of enhanced melanoma skin lesions through deep neural architectures. Information, 2022, 13(12), 583. 23. Prabu, S.; Jawali, N.; Sundar, K. J. A.; Sharvani, K.; Shanmukhanjali, G.; Nirmala, V. Indian Coin Detection and Recognition Using Deep Learning Algorithm. In 2022 6th Asian Conference on Artificial Intelligence Technology (ACAIT) (pp. 1–6). IEEE, 2022, December.
- 43. 16 DOI: 10.1201/9781003371380-2 Chapter 2 Trusted Location Information Verification Using Blockchain in Internet of Vehicles Ritesh Yaduwanshi and Sushil Kumar Jawaharlal Nehru University, New Delhi, India 2.1 Introduction In advanced vehicular adhoc networks (VANETs), a vehicle’s driver can connect to other vehicles’ drivers, to pedestrians, roadside units and others parts of the urban infrastructure. VANETs are an intelligent transportation system (ITS) that uses com- munication to reduce traffic congestion. VANETs can be driven by vehicle-to-road- side unit (RSU) and/or vehicle-to-vehicle communication [1]. A processing center (PC) can be an RSU (or a reliable vehicle with a predetermined position). Before giving directions to the vehicles inside its coverage area, the PC checks the commu- nication data. Location-based approaches and services are becoming commonplace in today’s wireless networks, so that location data verification has attracted a lot of coverage in recent years [1–9]. Generally vehicles get their location from GPS or GNSS, but the reported location information may be incorrect as the result of either faulty location data recording/forwarding technology, or deliberate misinformation. Undesirable network outcomes like insufficient toll payments, traffic congestion or traffic delays may result if the vehicle’s position information is not validated and the location inaccuracy is not recognized. In extreme situations, the lack of location
- 44. Trusted Location Information Verification ◾ 17 verification could result in disastrous events such as vehicle accidents. These various location verification systems (LVSs) that have been developed use a range of physi- cal layer signal parameters to evaluate the vehicle’s reported location information [2–11]. The constraint of all LVSs is that they typically perform well for the channel conditions that were considered throughout the design process [2], usually only working if all of the a priori channel data supplied to them is true. They can also only properly address the threat-model scenarios for which they were designed [12]. Because of these limitations, their real-world application is limited. The network can be disrupted simply by vehicles falsifying their location with location-based access control protocols [10] or geographic routing protocols [11]). A malicious vehicle can also falsify its location to seriously impair other vehicles [12] and to enhance its own network capabilities [13]. Accuracy of stated locations in VANETs is therefore critical and necessitates the use of an LVS. (See [14, 15] for an overview of IEEE 1609.2 certificate revocation.) This paper is organized as follows: related work is covered in Section 2.2, Section 2.3 presents the system model, Section 2.4 discusses results, and conclusions are offered in Section 2.5. 2.2 Related Work Location verification systems are required to overcome the problem of location falsifi- cation in VANETs. There are two types of existing verification systems: infrastructure- basedandinfrastructure-less[5,6].Everylocationverificationoperationinthismethod involves four base stations. Each one counts the time it takes to issue a challenge to the appropriate node and receive an individual response. However, the large number of verification requests from automobiles creates a network bottleneck at the base sta- tion. The cost of deploying and maintaining infrastructure also rises, making VANET an expensive network An infrastructure-based approach is therefore not suitable for VANETs. Most infrastructures-less verification systems use various distance measuring techniques to safely and transparently approve location assertions. For example, in ref. [7], the solution places verifiers at certain sites, each with its own allowable distance. The primary constraint of these techniques is the usage of non-RF range technology, which in turn increases the cost of building these networks. [8] proposes a method for achieving location verification based solely on logic beacon reception. Autonomous location verification was implemented in [9] without assuming different trust levels of nodes. For location verification, [10] uses location-limited channels. 2.3 Trusted Location Information Verification Using Blockchain In this section, we present a blockchain-based location verification system model focusing on location region and location sharping, location sharing and verification methods, together with the required assumptions.
- 45. 18 ◾ Intelligent Data Analytics, IoT, and Blockchain 2.3.1 Assumptions Vehicle-to-everything (V2X) and vehicle-to-vehicle (V2V) communications [11] are expected, as is the capacity for automobiles to connect to the internet efficiently. All vehicles are assumed to have the essential equipment, which includes sensors, GPS and on-board units (OBUs). It is believed that the number of valid roadside units (RSUs) outnumbers the number of malicious RSUs. We suppose that a valid RSU constructs a genesis block based on local events to initiate the block chain. We presume that important event alerts are disseminated within a certain geographic region of interest (RoI). We presume that the crucial signals are not encrypted, and that any adjacent vehicles will be able to read them. We assume that 15 messages are required to validate the occurrence, and that the message is correct. 2.3.2 System Model Figure 2.1 shows the two stages of the system model according to their functions of location sharing and location verification. By disclosing the real location coordi- nates and the real location region during the location sharing step, our suggested technique ensures that only Owner vehicles can pass location verification. A full description of the procedure follows. Table 2.1 Location Verification for Internet of Vehicles Articles Filtering Cryptography Infra structure Verification Detection Hubauz et al. [13] — — Yes Verifiable multilateration — Galle et al. [18] — — No Data-processing model Errors explained Xiao et al. [19] — Digital signature Yes Signal analysis Statistical model Leinmuller et al. [7] — — No Trust model using sensors — Yan et al. [20] — — No Radar Movement history Song et al. [22] — Symmetric keys No Signal analysis Distance enlargement Ren et al. [21] Grid map — No Filtered data — Ren et al. [21] — — No Directional antenna —
- 46. Trusted Location Information Verification ◾ 19 2.3.3 Location Sharing When a Requester asks for the Owner vehicle’s location information, there are two possibilities: either the Owner vehicle has complete faith in the Requester vehicle and hence provides precise coordinates (xi, yi), or the Owner vehicle does not com- pletely trust the Requester vehicle, in which case the Owner vehicle returns the rectangle location region with location coordinates (xi, yi). The pseudo code for the location sharing method is shown in Algorithm 1. The Owner vehicle will initially transmit the public key and session key by asymmetrically encrypting it, computing Res ← SE(ksession, xi||yi) [23, 24, 25, 26, 27] to encrypt the exact position: Res ← conRes||ASE(Pubo, ksession). Finally, the session key ksession is retrieved using the Requester’s private key to decrypt the Owner vehicle’s location coordinates, i.e., xi′||yi′ ← SD(ksession, conRes). The privacy-preserving process for location coordinates is essential when the Owner vehicle does not have total trust in the Requester vehicle (xi, yi). Through the underchain channel, the Owner vehicle computes fuzRes ← Enc(ksession, ciphi||borInfo‖nodesx|| nodesy) and provides Res ← fuzRes‖ASE(Pubr, ksession) to the Requester vehicle. Finally, using the session key ksession produced by the Requester vehicle’s private key Prir, the Owner vehicle’s location information is decrypted. The Requester vehicle obtains the privacy-preserving location area because the border plaintext information {xid1′, xid2′, yid3′, yid4′} is present in borInfo′. The remainder of fuzRes is used during the Requester vehicle’s location verification. The location sharing operation for the Owner vehicle and Requester vehicle is now complete. Figure 2.1 System model of block chain location verification in internet of vehicles.
- 47. 20 ◾ Intelligent Data Analytics, IoT, and Blockchain 2.3.4 Location Verification If the Requester vehicle confirms the Owner vehicle’s ith location information dur- ing location sharing, location verification takes place in two steps. The pseudocode for the location verification technique is shown in Algorithm 2. After the Requester decodes conRes, the Owner’s precise location coordinates li′ = (xi′, yi′) can be obtained if the Owner has complete trust in the Requester. The Requester then obtains the location record LRi created during the location record phase from the block chain. ALGORITHM 1 LOCATION SHARING Input: Privacy-preserving level n; Owner Vehicle’s Location li = (xi, yi); Owner vehicle’s public key Pubo Output: Shared Location Information (LS) a. Owner vehicle executes: b. If n = 0 then // trust level is maximum c. conRes ← SE(ksession, xi||yi); d. Res ← conRes||ASE (pubo, ksession); // exact coordinates are sent e. if N ≥ n ≥ 1 then // trust level is not maximum f. Find the border {xid1,xid2,yid3,yid4} in level n; //rectangular region sent g. borInfoid1 ← id1|| xid1|| ciphx id; h. borInfoid2 ← id2|| xid2|| ciphx id; i. borInfoid3 ← id3|| xid3|| ciphx id; j. borInfoid4 ← id4|| xid4|| ciphx id; k. borInfo ← borInfoid1|| borInfoid2|| borInfoid3|| borInfoid4; l. nodex ← {nodex x1, nodex x2,…..}; m. nodey ← {nodey y1, nodey y2,…..}; n. fuzRes ← Enc(ksession,ciphi,||borInfo||nodesx||nodesy); o. Res ← fuzRes||ASE(Pubr, ksession); p. End if q. Requester vehicle executes: r. If n = 0 then // trust level is maximum s. ||xi′yi′ ← SD(ksession,conRes); t. LS ← xi′|| yi′; u. Else if 1 ≤ n ≤ N then // trust level not maximum v. nodes′y ||nodes′x|| borInfo′|| ciphi′ ← SD(ksession,fuzRes) w. LS ← nodes′y || borInfo′|| nodes′x|| Ciphi′; x. End if y. Return LS
- 48. Trusted Location Information Verification ◾ 21 If Hash (xi′||yi′) = LRi·LIi·LHi, it shows that xi′ = xi and yi′ = yi. When the Owner vehicle has doubts about the Requester, ciphi′||borInfo′||nodes′x||nodes′ySD(ksess ion, fuzRes) is acquired after the Requester decodes fuzRes; nodes′x and nodes′y are needed nodes on the root node authentication path for recovering xTree and yTree. When the Owner has doubts about the Requester, ciphi′||borInfo′||nodes′x||nodes′y ← SD(ksession, fuzRes) is acquired after the Requester decodes fuzRes. The required nodes on the root node authentication path for recovering xTree and yTree are nodes′x and nodes′y. First, the Requester checks the received region boundary information borInfo′ for integrity. If genMT(Hash (borInfoid1′), Hash(borInfoid2′), nodes′x) = xTreeroot′ and genMT(Hash (borInfoid1′), Hash(borInfoid2′), nodes′y) = yTree- root′, the regional integrity verification is successful, indicating that the Owner has returned the proper region information borInfo′ = borInfo. If Hash(ciphi ′) = LRi·LIi·OpeHashi, (borInfoid1′·ciphx id1) ≤ (ciphi′·ciphx i) ≤ (borInfoid2′·ciphx id2), and (borInfoid3′·ciphy id3) ≤ (ciphi ′·Ciphy i) ≤ (borInfoid4′·ciphy id4), it shows that xid1 ≤ xi ≤ xid2 and yid3 ≤ yi ≤ yid4; location li is in the region surrounded by {xid1, xid2, yid3, yid4} that the Requester receives. The region verification is then complete. Finally, the Owner and Requester’s location verification operation is accomplished. The above algorithms have computational complexity of O(1) and O(1), respectively. With increasing N, the compute overhead of Algorithm 1 grows exponentially. Large plaintext spaces have a high computing overhead, although they can be implemented offline during the startup process. Furthermore, [28, 29, 30, 31, 32] Algorithm 1 is only used once throughout the entire procedure. The computing overhead of Algorithm 2 increases as plaintext space grows, with the value of N having no effect. In the location record phase, it also has a low computing overhead. The two ele- ments of N and plaintext space have essentially no effect on the processing cost of Algorithm 1. For Owners, the location sharing phase takes less time. Algorithm 4’s calculation overhead is approximately linear with N’s size. The plaintext space size has no bearing on it. Because the location verification phase only involves a hash operation, the computing overhead is minimal. ALGORITHM 2 LOCATION VERIFICATION Input: Location record in the Blockchain (LR); session key Ksession; Location sharing information (LS) Output: Boolean variable (LV) a. Initialize LV ← False b. If n = 0 then // trust level is maximum c. xi′||yi′ ← LS;
- 49. 22 ◾ Intelligent Data Analytics, IoT, and Blockchain 2.4 Results and Simulation 2.4.1 Location Leakage The likelihood of location leakage for each graph, as shown in Figure 2.2, is 0.075–0.0125 for distributed architecture and 0.125 for centralized architecture. Our suggested architecture can lower the probability to around 0.03%, significantly increasing location privacy safeguarding capability. 2.4.2 Channel Capacity Utilization The system adopts a cooperative approach, requiring message transmission between cars in close proximity. We looked at how often the system uses a wireless com- munication channel and how many messages are transferred between vehicles. We employed a 152 B packet payload that comprised request information as well as location information for the questioned node. The average channel utilization is shown in Figure 2.3. 2.4.3 Message Delivery Success Rate Single-hop delivery should allow vehicles within the same communication range to exchange messages. However, because of moving barriers, this may not always be the case. Figure 2.4 shows delivery of a message using a direct single-hop strategy to d. If Hash(xi′||yi′)LRi·Lli·LHi then e. LV ← True; f. End if g. Else if 1 ≤ n ≤ N then // trust level is not maximum h. ciphi′||borInfo′||nodes′x||nodes′y ← borInfo′; i. borInfo ← borInfoid1|| borInfoid2|| borInfoid3|| borInfoid4; j. xTreeroot′ ← genMT(Hash(broinfoid1′), Hash(broinfoid2′), nodes′x); k. yTreeroot′ ← genMT(Hash(broinfoid1′), Hash(broinfoid2′), nodes′y); l. If xTreeroot′ = xTreeroot and yTreeroot = yTreeroot then m. If Hash(ciphi′) = LRi·LIi·OpeHashi and borInfoid1′·ciphid1′ ciphi′·ciphi x borInfoid2′·ciphid2 x and borInfoid3′·ciphid3′ ciphi′·ciphi y and borInfoid4′·ciphid4 y n. LV ← True; o. End If p. End If q. End If r. Return LV
- 50. Trusted Location Information Verification ◾ 23 Figure 2.2 Location leakage. Figure 2.3 Channel capacity utilization.
- 51. 24 ◾ Intelligent Data Analytics, IoT, and Blockchain one that incorporated location verification and NLOS condition information. The sender can assess whether it can forward the message directly or whether it needs the help of other nodes by knowing the destination node. The results reveal that the delivery success rate has improved, as has the influence of moving impediments on direct messaging. 2.4.4 Processing Time The usual processing time from generation to verification response is depicted in Figure 2.5. A verification request typically takes less than 200 milliseconds to pro- cess, depending on vehicle density. 2.4.5 Security Attack Resilience We included malicious vehicular nodes in our simulations to evaluate our model’s security mechanisms. The malicious vehicular nodes carried out a variety of attacks that might disrupt the protocol. The results (see Figure 2.6) indicated that the tech- nique was not vulnerable to attacks. Malicious nodes accounted for between 25% and 75% of all nodes. These safeguards helped protect our protocol against the majority of the threats discovered. Figure 2.4 Message delivery success rate.
- 52. Trusted Location Information Verification ◾ 25 Figure 2.5 Processing time. Figure 2.6 Security attack resilience.
- 53. 26 ◾ Intelligent Data Analytics, IoT, and Blockchain 2.5 Conclusion This study investigates the decentralized architecture of an internet of vehicles based on block chain technology, and proposes a system model that includes block chain setup, vehicle registration, SBMs upload, and block chain record. Centralization and trustworthiness issues can be efficiently addressed by using a block chain-based VANET. In our proposed system model, there is no third central entity. The hash of SBMs is stored in block chain, which ensures SBM integrity while also speeding up data processing. The identity is then partitioned into more than k sub-identities in a blockchain-based internet of vehicles to ensure vehicle identity privacy, which will be updated on a regular basis using dynamic threshold encryption. The results of the experiments showed that our block chain-based internet of vehicles was very efficient in terms of system time and privacy protection. References 1. Kumar, S., Dohare, U., Kumar, K., Dora, D. P., Qureshi, K. N., Kharel, R. (2018). Cybersecurity measures for geocasting in vehicular cyber physical system environments. IEEE Internet of Things Journal, 6(4), 5916–5926. 2. Kumar, S., Singh, K., Kumar, S., Kaiwartya, O., Cao, Y., Zhou, H. (2019). Delimitated anti jammer scheme for Internet of vehicle: Machine learning based security approach. IEEE Access, 7, 113311–113323. 3. Kaiwartya, O., Cao, Y., Lloret, J., Kumar, S., Aslam, N., Kharel, R., … Shah, R. R. (2018). Geometry-based localization for GPS outage in vehicular cyber physical systems. IEEE Transactions on Vehicular Technology, 67(5), 3800–3812. 4. Kasana, R., Kumar, S., Kaiwartya, O., Yan, W., Cao, Y., Abdullah, A. H. (2017). Location error resilient geographical routing for vehicular ad-hoc networks. IET Intelligent Transport Systems, 11(8), 450–458. 5. Malaney, R. A. (2004, November). A location enabled wireless security system. In IEEE Global Telecommunications Conference, 2004. GLOBECOM’04. (Vol. 4, pp. 2196– 2200). IEEE. 6. Malandrino, F., Borgiattino, C., Casetti, C., Chiasserini, C. F., Fiore, M., Sadao, R. (2014). Verification and inference of positions in vehicular networks through anonymous beaconing. IEEE Transactions on Mobile Computing, 13(10), 2415–2428. 7. Leinmüller, T., Schoch, E., Kargl, F., Maihöfer, C. (2005, July). Influence of falsified position data on geographic ad-hoc routing. In EuropeanWorkshop on Security in Ad-hoc and Sensor Networks (pp. 102–112). Springer, Berlin, Heidelberg. 8. Čapkun, S., Čagalj, M., Karame, G., Tippenhauer, N. O. (2010). Integrity regions: authentication through presence in wireless networks. IEEE Transactions on Mobile Computing, 9(11), 1608–1621. 9. Raya, M., Hubaux, J. P. (2007). Securing vehicular ad hoc networks. Journal of Computer Security, 15(1), 39–68. 10. Yu, B., Xu, C. Z., Xiao, B. (2013). Detecting sybil attacks in VANETs. Journal of Parallel and Distributed Computing, 73(6), 746–756. 11. Zhang, T., Delgrossi, L. (2012). Vehicle safety communications: protocols, security, and privacy. John Wiley Sons.
- 54. Trusted Location Information Verification ◾ 27 12. Čapkun, S., Buttyán, L., Hubaux, J. P. (2003, October). SECTOR: secure tracking of node encounters in multi-hop wireless networks. In Proceedings of the 1st ACM Workshop on Security of Ad hoc and Sensor Networks (pp. 21–32). 13. Hubaux, J. P., Capkun, S., Luo, J. (2004). The security and privacy of smart vehicles. IEEE Security Privacy, 2(3), 49–55. 14. Sastry, N., Shankar, U., Wagner, D. (2003, September). Secure verification of location claims. In Proceedings of the 2nd ACM Workshop on Wireless Security (pp. 1–10). 15. Vora, A., Nesterenko, M. (2006). Secure location verification using radio broadcast. IEEE Transactions on Dependable and Secure Computing, 3(4), 377–385. 16. Xue, X., Lin, N., Ding, J., Ji, Y. (2010). A trusted neighbor table based location verifica- tion for VANET Routing. 17. Bd, S. (2002). SP, and WHC. In Talking to strangers: Authentication in adhoc wireless net- works, in Symposium on Network and Distributed Systems Security (NDSS’02). 18. Golle, P., Greene, D., Staddon, J. (2004, October). Detecting and correcting malicious data in VANETs. In Proceedings of the 1st ACM International Workshop on Vehicular Ad hoc Networks (pp. 29–37). 19. Xiao, B., Yu, B., Gao, C. (2006, September). Detection and localization of sybil nodes in VANETs. In Proceedings of the 2006 Workshop on Dependability Issues in Wireless Ad hoc Networks and Sensor Networks (pp. 1–8). 20. Yan, G., Olariu, S., Weigle, M. C. (2008). Providing VANET security through active position detection. Computer Communications, 31(12), 2883–2897. 21. Ren, Z., Li, W., Yang, Q. (2009, October). Location verification for VANETs routing. In 2009 IEEE International Conference on Wireless and Mobile Computing, Networking and Communications (pp. 141–146). IEEE. 22. Song, J. H., Wong, V. W., Leung, V. C. (2008, December). Secure location veri- fication for vehicular ad-hoc networks. In IEEE GLOBECOM 2008-2008 IEEE Global Telecommunications Conference (pp. 1–5). IEEE. 23. Yan, G., Chen, X., Olariu, S. (2009, October). Providing VANET position integrity through filtering. In 2009 12th International IEEE Conference on Intelligent Transportation Systems (pp. 1–6). IEEE. 24. Bucci, G., Ciancetta, F., Fiorucci, E., Fioravanti, A., Prudenzi, A., Mari, S. (2019, September). Challenge and future trends of distributed measurement systems based on Blockchain technology in the European context. In 2019 IEEE 10th International Workshop on Applied Measurements for Power Systems (AMPS) (pp. 1–6). IEEE. 25. Joy, J., Gerla, M. (2017, July). Internet of vehicles and autonomous connected car-pri- vacy and security issues. In 2017 26th International Conference on Computer Communication and Networks. 26. Joy, J., Cusack, G., Gerla, M. (2017, October). Poster: time analysis of the feasibility of vehicular blocktrees. In Proceedings of the 3rd Workshop on Experiences with the Design and Implementation of Smart Objects (pp. 25–26). 27. Dorri, A., Kanhere, S. S., Jurdak, R. (2017, April). Towards an optimized blockchain for IoT. In 2017 IEEE/ACM Second International Conference on Internet-of-Things Design and Implementation (IoTDI) (pp. 173–178). IEEE. 28. Sharma, P. K., Moon, S. Y., Park, J. H. (2017). Block-VN: A distributed blockchain based vehicular network architecture in smart city. Journal of Information Processing Systems, 13(1), 184–195. 29. Lei, A., Cruickshank, H., Cao, Y., Asuquo, P., Ogah, C. P. A., Sun, Z. (2017). Blockchain- based dynamic key management for heterogeneous intelligent transportation systems. IEEE Internet of Things Journal, 4(6), 1832–1843.
- 55. Random documents with unrelated content Scribd suggests to you:
- 56. before his father and uncle had ever gone out to the colony. He was here, and that father and uncle were dead; here, and on the way to what was undoubtedly his own property; a property to which no one could dispute his right, since George Ritherdon, his uncle, had been the only other heir his father had ever had. Yet, even as the animal which bore him continued to pace along amid all the rich tropical vegetation around them; even, too, as the yellow-headed parrots and the curassows chattered above his head and the monkeys leapt from branch to branch, he mused as to whether he was doing a wise thing in progressing towards Desolada- -the place where he was born, as he reflected with a strange feeling of incredulity in his mind. For suppose, he thought to himself, that when I get to it I find it shut up or in the occupation of some other settler--what am I to do then? How explain my appearance on the scene? I cannot very well ride up to the house on this animal and summon the garrison to surrender, like some knight-errant of old, and I can't stand parleying on the steps explaining who I am. I believe I have gone the wrong way to work after all! I ought to have gone and seen the Governor or the Chief Justice, or taken some advice, after stating who I was. Or Mr. Spranger! Confound it, why did I not present that letter of introduction to him before starting off here? The latter gentleman was a well-known planter and merchant living on the south side of Belize, to whom Julian had been furnished with a letter of introduction by a retired post-captain whom he had run against in London prior to his departure, and with whom he had dined at a Service Club. And this officer had given him so flattering an account of Mr. Spranger's hospitality, as well as the prominent position which that personage held in the little capital, that he now regretted considerably that he had not availed himself of the chance which had come in his way. More especially he regretted it, too, when there happened to come into his recollection the fact that the gallant sailor had stated with much enthusiasm--after dinner--that
- 57. Beatrix Spranger, the planter's daughter, was without doubt the prettiest as well as the nicest girl in the whole colony. However, he comforted himself with the reflection that the journey which he was now taking might easily serve as one of inspection simply, and that, as there was no particular hurry, he could return to Belize and then, before making any absolute claim upon his father's estate, take the advice of the most important people in the town. All of which, he said to himself, I ought to have thought of before and decided upon. However, it doesn't matter! A week hence will do just as well as now, and, meanwhile, I shall have had a look at the place which must undoubtedly belong to me. As he arrived at this conclusion, the mustang emerged from the forest-like copse they had been passing through, and ahead of him he saw, upon the flat plain, a little settlement or village. Which, thought Julian, must be All Pines. Especially as over there are the queer-shaped mountains called the 'Cockscomb,' of which the negro told me. Then he began to consider the advisability of finding accommodation at this place for a day or so while he made that inspection of the estate and residence of Desolada which he had on his ride decided upon. All Pines, to which he now drew very near, presented but a bare and straggling appearance, and that not a particularly flourishing one either. A factory fallen quite into disuse was passed by Julian as he approached the village; while although his eyes were able to see that, on its outskirts, there was more than one large sugar estate, the place itself was a poor one. Yet there was here that which the traveller finds everywhere, no matter to what part of the world he directs his footsteps and no matter how small the place he arrives at may be--an inn. An inn, outside which there were standing four or
- 58. five saddled mules and mustangs, and one fairly good-looking horse in excellent condition. A horse, however, that a person used to such animals might consider as showing rather more of the hinder white of its eye than was desirable, and which twitched its small, delicate ears in a manner equally suspicious. There seemed very little sign of life about this inn in spite of these animals, however, as Julian made his way into it, after tying up his own mustang to a nail in a tree--since a dog asleep outside in the sun and a negro asleep inside in what might be, and probably was, termed the entrance hall, scarcely furnished such signs. All the same, he heard voices, and pretty loud ones too, in some room close at hand, as well as something else, also--a sound which seemed familiar enough to his ears; a sound that he--who had been all over the world more than once as a sailor--had heard in diverse places. In Port Said to wit, in Shanghai, San Francisco, Lisbon, and Monte Carlo. The hum of a wheel, the click and rattle of a ball against brass, and then a soft voice--surely it was a woman's!--murmuring a number, a colour, a chance! So, so! said Julian to himself, Madame la Roulette, and here, too. Ah! well, madame is everywhere; why shouldn't she favour this place as well as all others that she can force her way into? Then he pushed open a swing door to his right, a door covered with cocoanut matting nailed on to it, perhaps to keep the place cool, perhaps to deaden sound--the sound of Madame la Roulette's clicking jaws--though surely this was scarcely necessary in such an out-of-the-way spot, and entered the room whence the noise proceeded. The place was darkened by matting and Persians; again, perhaps, to exclude the heat or deaden sound; and was, indeed, so dark that, until his eyes became accustomed to the dull gloom of the room-- vast and sparsely furnished--he could scarcely discern what was in it. He was, however, able to perceive the forms of four or five men
- 59. seated round a table, to see coins glittering on it; and a girl at the head of the table (so dark that, doubtless, she was of usual mixed Spanish and Indian blood common to the colony) who was acting as croupier--a girl in whose hair was an oleander flower that gleamed like a star in the general duskiness of her surroundings. While, as he gazed, she twirled the wheel, murmuring softly: Plank it down before it is too late, as well as, Make your game, and spun the ball; while, a moment later, she flung out pieces of gold and silver to right and left of her and raked in similar pieces, also from right and left of her. But the sordid, dusty room, across which the motes glanced in the single ray of sunshine that stole in and streamed across the table, was not--it need scarcely be said--a prototype of the gilded palace that smiles over the blue waters of the Mediterranean, nor of the great gambling chambers in the ancient streets behind the Cathedral in Lisbon, nor of the white and airy saloons of San Francisco-- instead, it was mean, dusty, and dirty, while over it there was the fœtid, sickly, tropical atmosphere that pervades places to which neither light nor constant air is often admitted. Himself unseen for the moment--since, as he entered the room, a wrangle had suddenly sprung up among all at the table over the disputed ownership of a certain stake--he stared in amazement into the gloomy den. Yet that amazement was not occasioned by the place itself (he had seen worse, or at least as bad, in other lands), but by the face of a man who was seated behind the half-caste girl acting as croupier, evidently under his directions. Where had he seen that face, or one like it, before? That was what he was asking himself now; that was what was causing his amazement! Where? Where? For the features were known to him--the face was familiar, some trick or turn in it was not strange.
- 60. Where had he done so, and what did it mean? Almost he was appalled, dismayed, at the sight of that face. The nose straight, the eyes full and clear, the chin clear cut; nothing in it unfamiliar to him except a certain cruel, determined look that he did not recognise. The dispute waxed stronger between the gamblers; the half-caste girl laughed and chattered like one of the monkeys outside in the woods, and beat the table more than once with her lithe, sinuous hand and summoned them to put down fresh stakes, to recommence the game; the men squabbled and wrangled between themselves, and one pointed significantly to his blouse--open at the breast; so significantly, indeed, that none who saw the action could doubt what there was inside that blouse, lying ready to his right hand. That action of the man--a little wizened fellow, himself half Spaniard, half Indian, with perhaps a drop or two of the tar-bucket also in his veins--brought things to an end, to a climax. For the other man whose face was puzzling Julian Ritherdon's brain, and puzzling him with a bewilderment that was almost weird and uncanny, suddenly sprang up from beside, or rather behind, the girl croupier and cried-- Stop it! Cease, I say. It is you, Jaime, you who always makes these disputes. Come! I'll have no more of it. And keep your hand from the pistol or---- But his threat was ended by his action, which was to seize the man he had addressed by the scruff of his neck, after which he commenced to haul him towards the door. Then he--then all of them--saw the intruder, Julian Ritherdon, standing there by that door, looking at them calmly and unruffled--
- 61. calm and unruffled, that is to say, except for his bewilderment at the sight of the other man's face. They all saw him in a moment as they turned, and in a moment a fresh uproar, a new disturbance, arose; a disturbance that seemed to bode ominously for Julian. For, now, in each man's hands there was a revolver, drawn like lightning from the breast of each shirt or blouse. Who are you? What are you? all cried together, except the girl, who was busily sweeping up the gold and silver on the table into her pockets. Who? One of the constabulary from Belize? A spy! Shoot him! No, exclaimed the man who bore the features that so amazed Julian Ritherdon, no, this is not one of the constabulary; while, as he spoke, his eyes roved over the tropical naval clothes, or whites, in which the former was clad for coolness. Neither do I believe he is a spy. Yet, he continued, what are you doing here? Who are you? Neither their pistols nor their cries had any power to alarm Julian, who, young as he was, had already won the Egyptian medal and the Albert medal for saving life; wherefore, looking his interrogator calmly in the face, he said-- I am on a visit to the colony, and my name is Julian Ritherdon. Julian Ritherdon! the other exclaimed, Julian Ritherdon! and as he spoke the owner of that name could see the astonishment on all their faces. Julian Ritherdon, he repeated again. That is it. Doubtless you know it hereabouts. May I be so bold as to ask what yours is? The man gave a hard, dry laugh--a strange laugh it was, too; then he replied, Certainly you may. Especially as mine is by chance
- 62. much the same as your own. My name is Sebastian Leigh Ritherdon. What! Your name is Ritherdon? You a Ritherdon? Who in Heaven's name are you, then? I happen to be the owner of a property near here called Desolada. The owner, because I am the son of the late Mr. Ritherdon and of his wife, Isobel Leigh, who died after giving me birth! CHAPTER V. A HALF-BREED NAMED ZARA. To describe Julian as being startled--amazed--would not convey the actual state of mind into which the answer given by the man who said that his name was Sebastian Leigh Ritherdon, plunged him. It was indeed something more than that; something more resembling a shock of consternation which now took possession of him. What did it mean?--he asked himself, even as he stood face to face with that other bearer of the name of Ritherdon. What? And to this question he could find but one answer: his uncle in England must, for some reason--the reason being in all probability that his hatred for the deceit practised on him years ago had never really become extinguished--have invented the whole story. Yet, of what use such an invention! How could he hope that he, Julian, should
- 63. profit by such a fabrication, by such a falsehood; why should he have bidden him go forth to a distant country there to assert a claim which could never be substantiated? Then, even in that moment, while still he stood astounded before the other Ritherdon, there flashed into his mind a second thought, another supposition; the thought that George Ritherdon had been a madman. That was--must be--the solution. None but a madman would have conceived such a story. If it were untrue! Yet, now, he could not pursue this train of thought; he must postpone reflection for the time being; he had to act, to speak, to give some account of himself. As to who he was, who, bearing the name of Ritherdon, had suddenly appeared in the very spot where Ritherdon was such a well-known and, probably, such an influential name. I never knew, the man who had announced himself as being the heir of the late Mr. Ritherdon was saying now, that there were any other Ritherdons in existence except my late father and myself; except myself now since his death. And, he continued, it is a little strange, perhaps, that I should learn such to be the case here in Honduras. Is it not? As he spoke to Julian, both his tone and manner were such as would not have produced an unfavourable impression upon any one who was witness to them. At the gaming-table, when seated behind the half-caste girl, his appearance would have probably been considered by some as sinister, while, when he had fallen upon the disputatious gambler, and had commenced--very roughly to hustle him towards the door, he had presented the appearance of a hectoring bully. Also, his first address to Julian on discovering him in the room had been by no means one that promised well for the probable events of the next few moments. But now--now--his manner and whole bearing were in no way aggressive, even though
- 64. his words expressed that a certain doubt in his mind accompanied them. Surely, he continued, we must be connections of some sort. The presence of a Ritherdon in Honduras, within an hour's ride of my property, must be owing to something more than coincidence. It is owing to something more than coincidence, Julian replied, scorning to take refuge in an absolute falsehood, though acknowledging to himself that, in the position in which he now found himself--and until he could think matters out more clearly, as well as obtain some light on the strange circumstances in which he was suddenly involved--diplomacy if not evasion--a hateful word!--was necessary. More than coincidence. You may have heard of George Ritherdon, your uncle, who once lived here in the colony with your father. Yes, Sebastian Ritherdon answered, his eyes still on the other. Yes, I have heard my father speak of him. Yet, that was years ago. Nearly thirty, I think. Is he here, too? In the colony? No; he is dead. But I am his son. And, being on leave from my profession, which is that of an officer in her Majesty's navy, it has suited me to pay a visit to a place of which he had spoken so often. As he gave this answer, Julian was able to console himself with the reflection that, although there was evasion in it, at least there was no falsehood. For had he not always believed himself to be George Ritherdon's son until a month or so ago; had he not been brought up and entered for the navy as his son? Also, was he sure now that he was not his son? He had listened to a story from the dying man telling how he, Julian, had been kidnapped from his father's house, and how the latter had been left childless and desolate; yet now, when he was almost at the threshold of that house, he found himself face to face with a man, evidently well
- 65. known in all the district, who proclaimed himself to be the actual son--a man who also gave, with some distinctness in his tone, the name of Isobel Leigh as that of his mother. She Sebastian Ritherdon's mother! the woman who was, he had been told, his own mother: the woman who, dying in giving birth to her first son, could consequently have never been the mother of a second. Was it not well, therefore, that, as he had always been, so he should continue to be, certainly for the present, the son of George Ritherdon, and not of Charles? For, to proclaim himself here, in Honduras, as the offspring of the latter would be to bring down upon him, almost of a surety, the charge of being an impostor. I knew, exclaimed Sebastian, while in his look and manner there was expressed considerable cordiality; I knew we must be akin. I was certain of it. Even as you stood in that doorway, and as the ray of sunlight streamed across the room, I felt sure of it before you mentioned your name. Why? asked Julian surprised; perhaps, too, a little agitated. Why! Can you not understand? Not recognise why--at once? Man alive! We are alike! Alike! Alike! The words fell on Julian with startling force. Alike! Yes, so they were! They were alike. And in an instant it seemed as if some veil, some web had fallen away from his mental vision; as if he understood what had hitherto puzzled him. He understood his bewilderment as to where he had seen that face and those features before! For now he knew. He had seen them in the looking-glass! No doubt about the likeness! exclaimed one of the gamblers who had remained in the room, a listener to the conference; while the half-breed stared from first one face to the other with her large eyes wide open. No doubt about that. As much like brothers as cousins, I should say.
- 66. And the girl who (since Julian's intrusion, and since, also, she had discovered that it was not the constabulary from Belize who had suddenly raided their gambling den), had preserved a stolid silence-- glancing ever and anon with dusky eyes at each, muttered also that none who saw those two men together could doubt that they were kinsmen, or, as she termed it, parienti. Yes, Julian answered bewildered, almost stunned, as one thing after another seemed--with crushing force--to be sweeping away for ever all possibility of George Ritherdon's story having had any foundation in fact, any likelihood of being aught else but the chimera of a distraught brain; yes, I can perceive it. I--I--wondered where I had seen your face before, when I first entered the room. Now I know. And, Sebastian exclaimed, slapping his newly found kinsmen somewhat boisterously on the back, and we are cousins. So much the better! For my part I am heartily glad to meet a relation. Now-- come--let us be off to Desolada. You were on your way there, no doubt. Well! you shall have a cordial welcome. The best I can offer. You know that the Spaniards always call their house 'their guests' house.' And my house shall be yours. For as long as you like to make it so. You are very good, Julian said haltingly, feeling, too, that he was no longer master of himself, no longer possessed of all that ease which he had, until to-day, imagined himself to be in full possession of. Very good indeed. And what you say is the case. I was on my way--I--had a desire to see the place in which your and my father lived. You shall see it, you shall be most welcome. And, Sebastian continued, you will find it big enough. It is a vast rambling place, half wood, half brick, constructed originally by Spanish settlers, so that it is over a hundred years old. The name is a mournful one, yet it has always been retained. And once it was appropriate enough.
- 67. There was scarcely another dwelling near it for miles--as a matter of fact, there are hardly any now. The nearest, which is a place called 'La Superba,' is five miles farther on. They went out together now to the front of the inn--Julian observing that still the negro slept on in the entrance-hall and still the dog slept on in the sun outside--and here Sebastian, finding the good-looking horse, began to untether it, while Julian did the same for his mustang. They were the only two animals now left standing in the shade thrown by the house, since all the men--including he who had stayed last and listened to their conversation--were gone. The girl, however, still remained, and to her Sebastian spoke, bidding her make her way through the bypaths of the forest to Desolada and state that he and his guest were coming. Who is she? asked Julian, feeling that it was incumbent on him to evince some interest in this new-found cousin's affairs; while, as was not surprising, he really felt too dazed to heed much that was passing around him. The astonishment, the bewilderment that had fallen on him owing to the events of the last half-hour, the startling information he had received, all of which tended, if it did anything, to disprove every word that George Ritherdon had uttered prior to his death--were enough to daze a man of even cooler instincts than he possessed. She, said Sebastian, with a half laugh, a laugh in which contempt was strangely discernible, she, oh! she's a half-breed-- Spanish and native mixed--named Zara. She was born on our place and turns her hand to anything required, from milking the goats to superintending the negroes. She seems to know how to turn her hand to a roulette wheel also, Julian remarked, still endeavouring to frame some sentences which should pass muster for the ordinary courteous attention expected from a newly found relation, who had also, now, assumed the character of guest.
- 68. Yes, Sebastian answered. Yes, she can do that too. I suppose you were surprised at finding all the implements of a gambling room here! Yet, if you lived in the colony it would not seem so strange. We planters, especially in the wild parts, must have some amusement, even though it's illegal. Therefore, we meet three times a week at the inn, and the man who is willing to put down the most money takes the bank. It happened to me to-day. And, as in the case of most hot countries, said Julian, forcing himself to be interested, a servant is used for that portion of the game which necessitates exertion. I understand! In some tropical countries I have known, men bring their servants to deal for them at whist and mark their game. You have seen a great deal of the world as a sailor? the other asked, while they now wended their way through a thick mangrove wood in which the monkeys and parrots kept up such an incessant chattering that they could scarcely hear themselves talk. I have been round it three times, Julian replied; though, of course, sailor-like, I know the coast portions of different countries much better than I do any of the interiors. And I have never been farther away than New Orleans. My mother ca--my mother always wanted to go there and see it. Was she--your mother from New Orleans? Julian asked, on the alert at this moment, he hardly knew why. My mother. Oh! no. She was the daughter of Mr. Leigh, an English merchant at Belize. But, as you will discover, New Orleans means the world to us--we all want to go there sometimes.
- 69. CHAPTER VI. KNOWLEDGE IS NOT ALWAYS PROOF. If there was one desire more paramount than another in Julian's mind--as now they threaded a campeachy wood dotted here and there with clumps of cabbage palms while, all around, in the underbrush and pools, the Caribbean lily grew in thick and luxurious profusion--that desire was to be alone. To be able to reflect and to think uninterruptedly, and without being obliged at every moment to listen to his companion's flow of conversation--which was so unceasing that it seemed forced--as well as obliged to answer questions and to display an interest in all that was being said. Julian felt, perhaps, this desire the more strongly because, by now, he was gradually becoming able to collect himself, to adjust his thoughts and reflections and, thereby, to bring a more calm and clear insight to bear upon the discovery--so amazing and surprising-- which had come to his knowledge but an hour or so ago. If he were alone now, he told himself, if he could only get half-an-hour's entire and uninterrupted freedom for thought, he could, he felt sure, review the matter with coolness and judgment. Also, he could ponder over one or two things which, at this moment, struck him with a force they had not done at the time when they had fallen with stunning--because unexpected--force upon his brain. Things--namely words and statements--that might go far towards explaining, if not towards unravelling, much that had hitherto seemed inexplicable. Yet, all the same, he was obliged to confess to himself that one thing seemed absolutely incapable of explanation. That was, how this man could be the child of Charles Ritherdon, the late owner of the vast property through which they were now riding, if his brother George had been neither demented nor a liar. And that Sebastian should have invented his statement was obviously incredible for the plain and simple reasons that he had made it before several
- 70. witnesses, and that he was in full possession, as recognised heir, of all that the dead planter had left behind. It was impossible, however, that he could meditate--and, certainly, he could not follow any train of thought--amid the unfailing flow of conversation in which his companion indulged. That flow gave him the impression, as it must have given any other person who might by chance have overheard it, that it was conversation made for conversation's sake, or, in other words, made with a determination to preclude all reflection on Julian's part. From one thing to another this man, called Sebastian Ritherdon, wandered-- from the trade of the colony to its products and vegetation, to the climate, the melancholy and loneliness of life in the whole district, the absence of news and of excitement, the stagnation of everything except the power of making money by exportation. Then, when all these topics appeared to be thoroughly beaten out and exhausted, Sebastian Ritherdon recurred to a remark made during the earlier part of their ride, and said: So you have a letter of introduction to the Sprangers? Well! you should present it. Old Spranger is a pleasant, agreeable man, while as for Beatrix, his daughter, she is a beautiful girl. Wasted here, though. Is she? said Julian. Are there, then, no eligible men in British Honduras who could prevent a beautiful girl from failing in what every beautiful girl hopes to accomplish--namely getting well settled? Oh, yes! the other answered, and now it seemed to Julian as though in his tone there was something which spoke of disappointment, if not of regret, personal to the man himself. Oh, yes! There are such men among us. Men well-to-do, large owners of remunerative estates, capitalists employing a good deal of labour, and so forth. Only--only----
- 71. Only what? Well--oh! I don't know; perhaps we are not quite her class, her style. In England the Sprangers are somebody, I believe, and Beatrix is consequently rather difficult to please. At any rate I know she has rejected more than one good offer. She will never marry any colonist. Then, as Julian turned his eyes on Sebastian Ritherdon, he felt as sure as if the man had told him so himself that he was one of the rejected. I intend to present that letter of introduction, you know, he said a moment later. In fact I intended to do so from the first. Now, your description of Miss Spranger makes me the more eager. You may suit her, the other replied. I mean, of course, as a friend, a companion. You are a naval officer, consequently a gentleman in manners, a man of the world and of society. As for us, well, we may be gentlemen, too, only we don't, of course, know much about society manners. He paused a moment--it was indeed the longest pause he had made for some time; then he said, When do you propose to go to see them? I rather thought I would go back to Belize to-morrow, Julian answered. To-morrow! Yes. I--I--feel I ought not to be in the country and not present that letter. To-morrow! Sebastian Ritherdon said again. To-morrow! That won't give me much of your society. And I'm your cousin.
- 72. Oh! said Julian, forcing a smile, you will have plenty of that--of my society--I'm afraid. I have a long leave, and if you will have me, I will promise to weary you sufficiently before I finally depart. You will be tired enough of me ere then. To his surprise--since nothing that the other said (and not even the fact that the man was undoubtedly regarded by all who knew him as the son and heir of Mr. Ritherdon and was in absolute fact in full possession of the rights of such an heir) could make Julian believe that his presence was a welcome one--to his surprise, Sebastian Ritherdon greeted his remark with effusion. None who saw his smile, and the manner in which his face lit up, could have doubted that the other's promise to stay as his guest for a considerable time gave him the greatest pleasure. Then, suddenly, while he was telling Julian so, they emerged from one more glade, leaving behind them all the chattering members of the animal and feathered world, and came out into a small open plain which was in a full state of cultivation, while Julian observed a house, large, spacious and low before them. There is Desolada--the House of Desolation as my poor father used to call it, for some reason of his own--there is my property, to which you will always be welcome. His property! Julian thought, even as he gazed upon the mansion (for such it was); his property! And he had left England, had travelled thousands of miles to reach it, thinking that, instead, it was his. That he would find it awaiting an owner--perhaps in charge of some Government official, but still awaiting an owner--himself. Yet, now, how different all was from what he had imagined--how different! In England, on the voyage, the journey from New York to New Orleans, nay! until four hours ago, he thought that he would have but to tell his story after taking a hasty view of Desolada and its surroundings to prove that he was the son who had suddenly disappeared a day or so after his birth: to show that he was the
- 73. missing, kidnapped child. He would have but to proclaim himself and be acknowledged. But, lo! how changed all appeared now. There was no missing, kidnapped heir--there could not be if the man by his side had spoken the truth--and how could he have spoken untruthfully here, in this country, in this district, where a falsehood such as that statement would have been (if not capable of immediate and universal corroboration), was open to instant denial? There must be hundreds of people in the colony who had known Sebastian Ritherdon from his infancy; every one in the colony would have been acquainted with such a fact as the kidnapping of the wealthy Mr. Ritherdon's heir if it had ever taken place, and, in such circumstances, there could have been no Sebastian. Yet here he was by Julian's side escorting him to his own house, proclaiming himself the owner of that house and property. Surely it was impossible that the statement could be untrue! Yet, if true, who was he himself? What! What could he be but a man who had been used by his dying father as one who, by an imposture, might be made the instrument of a long-conceived desire for vengeance--a vengeance to be worked out by fraud? A man who would at once have been branded as an impostor had he but made the claim he had quitted England with the intention of making. Under the palms--which grew in groves and were used as shade- trees--beneath the umbrageous figs, through a garden in which the oleanders flowered luxuriously, and the plants and mignonette-trees perfumed deliciously the evening air, while flamboyants--bearing masses of scarlet, bloodlike flowers--allamandas, and temple-plants gave a brilliant colouring to the scene, they rode up to the steps of the house, around the whole of which there was a wooden balcony. Standing upon that balcony, which was made to traverse the vast mansion so that, no matter where the sun happened to be, it could be avoided, was a woman, smiling and waving her hand to Sebastian, although it seemed that, in the salutation, the newcomer
- 74. was included. A woman who, in the shadow which enveloped her, since now the sun had sunk away to the back, appeared so dark of complexion as to suggest that in her veins there ran the dark blood of Africa. Yet, a moment later, as Sebastian Ritherdon presented Julian to her, terming him a new-found cousin, the latter was able to perceive that the shadows of the coming tropical night had played tricks with him. In this woman's veins there ran no drop of black blood; instead, she was only a dark, handsome Creole--one who, in her day, must have been even more than handsome--must have possessed superb beauty. But that day had passed now, she evidently being near her fiftieth year, though the clear ivory complexion, the black curling hair, in which scarcely a grey streak was visible, the soft rounded features and the dark eyes, still full of lustre, proclaimed distinctly what her beauty must have been in long past days. Also, Julian noticed, as she held out a white slim hand and murmured some words of cordial welcome to him, that her figure, lithe and sinuous, was one that might have become a woman young enough to have been her daughter. Only--he thought--it was almost too lithe and sinuous: it reminded him too much of a tiger he had once stalked in India, and of how he had seen the striped body creeping in and out of the jungle. This is Madame Carmaux, Sebastian said to Julian, as the latter bowed before her, a relation of my late mother. She has been here many years--even before that mother died. And--she has been one to me as well as fulfilling all the duties of the lady of the house both for my father and, now, for myself. Then, after Julian had muttered some suitable words and had once more received a gracious smile from the owner of those dark eyes, Sebastian said, Now, you would like to make some kind of toilette, I suppose, before the evening meal. Come, I will show you
- 75. your room. And he led the way up the vast campeachy-wood staircase to the floor above. Tropical nights fall swiftly directly the sun has disappeared, as it had now done behind the still gilded crests of the Cockscomb range, and Julian, standing on his balcony after the other had left him and gazing out on all around, wondered what was to be the outcome of this visit to Honduras. He pondered, too, as he had pondered before, whether George Ritherdon had in truth been a madman or one who had plotted a strange scheme of revenge against his brother; a scheme which now could never be perfected. Or--for he mused on this also--had George Ritherdon spoken the truth, had Sebastian---- The current of his thoughts was broken, even as he arrived at this point, by hearing beneath him on the under balcony the voice of Sebastian speaking in tones low but clear and distinct--by hearing that voice say, as though in answer to another's question: Know--of course he must know! But knowledge is not always proof. CHAPTER VII. MADAME CARMAUX TAKES A NAP On that night when Sebastian Ritherdon escorted Julian once more up the great campeachy-wood staircase to the room allotted to him, he had extorted a promise from his guest that he would stay at least one day before breaking his visit by another to Sprangers.
- 76. For, he had said before, down in the vast dining-room--which would almost have served for a modern Continental hotel--and now said again ere he bid his cousin good-night, for what does one day matter? And, you know, you can return to Belize twice as fast as you came here. How so? asked Julian, while, as he spoke, his eyes were roaming round the great desolate corridors of the first floor, and he was, almost unknowingly to himself, peering down those corridors amid the shadows which the lamp that Sebastian carried scarcely served to illuminate. How so? Why, first, you know your road now. Then, next, I can mount you on a good swift trotting horse that will do the journey in a third of the time that mustang took to get you along. How ever did you become possessed of such a creature? We rarely see them here. I hired it from the man who kept the hotel. He said it was the proper thing to do the journey with. Proper thing, indeed! More proper to assist the bullocks and mules in transporting the mahogany and campeachy, or the fruits, from the interior to the coast. However, you shall have a good trotting Spanish horse to take you into Belize, and I'll send your creature back later. Then, after wishing each other good-night, Julian entered the room, Sebastian handing him the lamp he had carried upstairs to light the way. I can find my own way down again in the dark very well, the latter said. I ought to be able to do so in the house I was born in and have lived in all my life. Good-night. At last Julian was alone. Alone with some hours before him in which he could reflect and meditate on the occurrences of this eventful day.
- 77. He did now that which perhaps, every man, no matter how courageous he might have been, would have done in similar circumstances. He made a careful inspection of the room, looking into a large wardrobe which stood in the corner, and, it must be admitted, under the bed also; which, as is the case in most tropical climates, stood in the middle of the room, so that the mosquitoes that harboured in the whitewashed walls should have less opportunity of forcing their way through the gauze nets which protected the bed. Then, having completed this survey to his satisfaction, he put his hand into his breast and drew from a pocket inside his waistcoat that which, it may well be surmised, he was not very likely to be without here. This was an express revolver. That's all right, he said as, after a glance at the chambers, he laid it on the table by his side. You have been of use before, my friend, in other parts of the world and, although you are not likely to be wanted here, you don't take up much room. Now, he went on to himself, for a good long think, as the paymaster of the Mongoose always used to say before he fell asleep in the wardroom and drove everybody else out of it with his snores. Only, first there are one or two other little things to be done. Whereon he walked out on to the balcony--the windows of course being open--and gave a long and searching glance around, above, and below him. Below, to where was the veranda of the lower or ground floor, with, standing about, two or three Singapore chairs covered with chintz, a small table and, upon it, a bottle of spirits and some glasses as well as a large carafe of water. All these things were perfectly visible because, from the room beneath him, there streamed out a strong light from the oil lamp which stood on the table within that room, while, even though such had not been the case, Julian was perfectly well aware that they were there. He and Sebastian had sat in those chairs for more than an hour talking after the evening meal, while Madame Carmaux, whose other
- 78. name he learnt was Miriam, had sat in another, perusing by the light of the lamp the Belize Advertiser. Yet, now and again, it had seemed to Julian as though, while those dark eyes had been fixed on the sheet, their owner's attention had been otherwise occupied, or else that she read very slowly. For once, when he had been giving a very guarded description of George Ritherdon's life in England during the last few years, he had seen them rest momentarily upon his face, and then be quickly withdrawn. Also, he had observed, the newspaper had never been turned once. Now, he said again to himself, now, let us think it all out and come to some decision as to what it all means. Let us see. Let me go over everything that has happened since I pulled up outside that inn--or gambling house! He was, perhaps, a little more methodical than most young men; the habit being doubtless born of many examinations at Greenwich, of a long course in H.M.S. Excellent, and, possibly, of the fact that he had done what sailors call a lot of logging in his time, both as watchkeeper and when in command of a destroyer. Therefore, he drew from his pocket a rather large, but somewhat unbusinesslike- looking pocketbook--since it was bound in crushed morocco and had its leaves gilt-edged--and, ruthlessly tearing out a sheet of paper, he withdrew the pencil from its place and prepared to make notes. No orders as to 'lights out,' he muttered to himself before beginning. I suppose I may sit up as long as I like. Then, after a few moments' reflection, he jotted down: S. didn't seem astonished to see me. (Qy?) Ought to have done so, if I came as a surprise to him. Can't ever have heard of me before. Consequently it was a surprise. Said who he was, and was particularly careful to say who his mother was, viz. I. S. R. (Qy?) Isn't that odd? Known many people who tell you who their father was. Never knew 'em lug in their mother's name, though, except
- 79. when very swagger. Says Madame Carmaux relative of his mother, yet Isobel Leigh was daughter of English planter. C's not a full-bred Englishwoman, and her name's French. That's nothing, though. Perhaps married a Frenchman. These little notes--which filled the detached sheet of the ornamental pocketbook--being written down, Julian, before taking another, sat back in his chair to ponder; yet his musings were not satisfactory, and, indeed, did not tend to enlighten him very much, which, as a matter of fact, they were not very likely to do. He must be the right man, after all, and I must be the wrong one, he said to himself. It is impossible the thing can be otherwise. A child kidnapped would make such a sensation in a place like this that the affair would furnish gossip for the next fifty years. Also, if a child was kidnapped, how on earth has this man grown up here and now inherited the property? If I was actually the child I certainly didn't grow up here, and if he was the child and did grow up here then there was no kidnapping. Indeed, by the time that Julian had arrived at this rather complicated result, he began to feel that his brain was getting into a whirl, and he came to a hasty resolution. That resolution was that he would abandon this business altogether; that, on the next day but one, he would go to Belize and pay his visit to the Sprangers, while, when that visit was concluded, he would, instead of returning to Desolada, set out on his return journey to England. Even though my uncle--if he was my uncle and not my father-- spoke the truth and told everything exactly as it occurred, how is it to be proved? How can any legal power on earth dispossess a man who has been brought up here from his infancy, in favour of one who comes without any evidence in his favour, since that certificate of my baptism in New Orleans, although it states me to be the son of the late owner of this place, cannot be substantiated? Any man might have taken any child and had such an entry as that made. And
- 80. if he--he my uncle, or my father--could conceive such a scheme as he revealed to me--or such a scheme as he did not reveal to me-- then, the entry at New Orleans would not present much difficulty to one like him. It is proof--proof that it be---- He stopped in his meditations--stopped, wondering where he had heard something said about proof before on this evening. Then, in a moment, he recalled the almost whispered words; the words that in absolute fact were whispered from the balcony below, before he went down to take his seat at the supper table; the utterance of Sebastian: Know--of course he must know. But knowledge is not always proof. How strange it was, he thought, that, while he had been indulging in his musings, jotting down his little facts on the sheet of paper, he should have forgotten those words. Knowledge is not always proof. What knowledge? Whose? Whose could it be but his! Whose knowledge that was not proof had Sebastian referred to? Then again, in a moment--again suddenly--he came to another determination, another resolve. He did possess some knowledge that this man, Sebastian could not dispute--for it would have been folly to imagine he had been speaking of any one else but him--though he had no proof. So be it, only, now, he would endeavour to discover a proof that should justify such knowledge. He would not slink away from the colony until he had exhausted every attempt to discover that proof. If it was to be found he would find it. Perhaps, after all, his uncle was his uncle, perhaps that uncle had undoubtedly uttered the truth. He rose now, preparing to go to bed, and as he did so a slight breeze rattled the slats of the green persianas, or, as they are called in England, Venetian blinds--a breeze that in tropical land often rises
- 81. as the night goes on. It was a cooling pleasant one, and he remembered that he had heard it rustling the slats before, when he was engaged in making his notes. Yet, now, regarding those green strips of wood, he felt a little astonished at what he saw. He had carefully let the blinds of both windows down and turned the laths so that neither bats nor moths, nor any of the flying insect world which are the curse of the tropics at night, should force their way in, attracted by the flame of the lamp; but now, one of those laths was turned--turned, so that, instead of being downwards and forming with the others a compact screen from the outside, it was in a flat or horizontal position, leaving an open space of an inch between it and the one above and the next below. A slat that was above five feet from the bottom of the blind. He stood there regarding it for a moment; then, dropping the revolver into his pocket, he went towards the window and with his finger and thumb put back the lath into the position he had originally placed it, feeling as he did so that it did not move smoothly, but, instead, a little stiffly. There has been no wind coming up from the sea that would do that, he reflected, and, if it had come, then it would have turned more than one. I wonder whether, and now he felt a slight sensation of creepiness coming over him, if I had raised my eyes as I sat writing, I should have met another pair of eyes looking in on me. Very likely. The turning of that one lath made a peep-hole. He pulled the blind up now without any attempt at concealing the noise it caused--that well-known clatter made by such blinds as they are hastily drawn up--and walked out on to the long balcony and peered over on to the one beneath, seeing that Madame Carmaux was asleep in the wicker chair which she had sat in during the evening, and that the newspaper lay in her lap. He saw, too, that Sebastian Ritherdon was also sitting in his chair, but that, aroused by
- 82. the noise of the blind, he had bent his body backwards over the veranda rail and, with upturned face, was regarding the spot at which Julian might be expected to appear. Not gone to bed, yet, old fellow, he called out now, on seeing the other lean over the balcony rail; while Julian observed that Madame Carmaux opened her eyes with a dazzled look--the look which those have on their faces who are suddenly startled out of a light nap. And for some reason--since he was growing suspicious--he believed that look to have been assumed as well as the slumber which had apparently preceded it. CHAPTER VIII. A MIDNIGHT VISITOR Not yet, Julian called down in answer to the other's remark, though I am going directly. Only it is so hot. I hope I am not disturbing the house. Not at all. Do what you like. We often sit here till long after midnight, since it is the only cool time of the twenty-four hours. Will you come down again and join us? No, if you'll excuse me. I'll take a turn or two here and then go to bed.
- 83. Whereon as he spoke, he began to walk up and down the balcony. It ran (as has been said of the lower one on which Sebastian and Madame Carmaux were seated) round the whole of the house, so that, had Julian desired to do so, he could have commenced a tour of the building which, by being continued, would eventually have brought him back to the spot where he now was. He contented himself, however, with commencing to walk towards the right-hand corner of the great rambling mansion, proceeding as far upon it as led to where the balcony turned at the angle, then, after a glance down its--at that place--darkened length, he retraced his steps, meaning to proceed to the opposite or left-hand corner. Doing so, however, and coming thus in front of his bedroom window, from which, since the blind was up, the light of his lamp streamed out on to the broad wooden floor of the balcony, he saw lying at his feet a small object which formed a patch of colour on the dark boards. A patch which was of a pale roseate hue, the thing being, indeed, a little spray, now dry and faded, of the oleander flower. And he knew, felt sure, where he had seen that spray before. I know now, he said to himself, who turned the slat--who stood outside my window looking in on me. Picking up the withered thing, he, nevertheless, continued his stroll along the balcony until he arrived at the left angle of the house, when he was able to glance down the whole of that side of it, this being as much in the dark and unrelieved by any light from within as the corresponding right side had been. Unrelieved, that is, by any light except the gleam of the great stars which here glisten with an incandescent whiteness; and in that gleam he saw sitting on the floor of the balcony--her back against the wall, her arms over her knees and her head sunk on those arms--the half-caste girl, Zara, the croupier of the gambling-table to which Sebastian had supplied the bank that morning at All Pines.
- 84. You have dropped this flower from your hair, he said, tossing it lightly down to her, while she turned up her dark, dusky eyes at him and, picking up the withered spray, tossed it in her turn contemptuously over the balcony. But she said nothing and, a moment later, let her head droop once more towards her arms. Do you pass the night here? he said now. Surely it is not wholesome to keep out in open air like this. I sit here often, she replied, before going to bed in my room behind. The rooms are too warm. I disturb no one. For a moment he felt disposed to say that it would disturb him if she should again take it into her head to turn his blinds, but, on second considerations, he held his peace. To know a thing and not to divulge one's knowledge is, he reflected, sometimes to possess a secret--a clue--a warning worth having; to possess, indeed, something that may be of use to us in the future if not now, while, for the rest--well! the returning of the spray to her had, doubtless, informed the girl sufficiently that he was acquainted with the fact of how she had been outside his window, and that it was she who had opened his blind wide enough to allow her to peer in on him. Good-night, he said, turning away. Good-night, and without waiting to hear whether she returned the greeting or not, he went back to the bedroom. Yet, before he entered it, he bent over the balcony and called down another good-night to Sebastian, who, he noticed, had now been deserted by Madame Carmaux. For some considerable time after this he walked about his room; long enough, indeed, to give Sebastian the idea that he was preparing for bed, then, although he had removed none of his clothing except his boots, he put out the lamp. If the young lady is desirous of observing me again, he reflected, she can do so. Yet if she does, it will not be without my
- 85. knowing it. And if she should pay me another visit--why, we shall see. But, all the same, and because he thought it not at all unlikely that some other visitor than the girl might make her way, not only to the blind itself but even to the room, he laid his right arm along the table so that his fingers were touching the revolver that he had now placed on that table. I haven't taken countless middle watches for nothing in my time, he said to himself; another won't hurt me. If I do drop asleep, I imagine I shall wake up pretty easily. He was on the alert now, and not only on the alert as to any one who might be disposed to pay him a nocturnal visit, but, also, mentally wary as to what might be the truth concerning Sebastian Ritherdon and himself. For, strange to say, there was a singular revulsion of feeling going on in his mind at this time; strange because, at present, scarcely anything of considerable importance, scarcely anything sufficiently tangible, had occurred to produce this new conviction that Sebastian's story was untrue, and that the other story told by his uncle before his death was the right one. All the same, the conviction was growing in his mind; growing steadily, although perhaps without any just reason or cause for its growth. Meanwhile, his ears now told him that, although Madame Carmaux was absent when he glanced over the balcony to wish Sebastian that last greeting, she undoubtedly had not gone to bed. From below, in the intense stillness of the tropic night--a stillness broken only occasionally by the cry of some bird from the plantation beyond the cultivated gardens, he heard the soft luscious tones of the woman herself--and those who are familiar with the tones of southern women will recall how luscious the murmur can be; he heard, too, the deeper notes of the man. Yet what they said to each other in subdued whispers was unintelligible to him; beyond a word here and there nothing reached his ears.
- 86. With the feeling of conviction growing stronger and stronger in his mind that there was some deception about the whole affair--that, plausible as Sebastian's possession of all which the dead man had left behind appeared; plausible, too, as was his undoubted position here and had been from his very earliest days, Julian would have given much now to overhear their conversation--a conversation which, he felt certain, in spite of it taking place thirty feet below where he was supposed to be by now asleep, related to his appearance on the scene. Would it be possible? Could he in any way manage to thus overhear it? If he were nearer to the persianas, his ear close to the slats, his head placed down low, close to the boards of the room and of the balcony as well--what might not be overheard? Thinking thus, he resolved to make the attempt, even while he told himself that in no other circumstances would he--a gentleman, a man of honour--resort to such a scheme of prying interference. But-- for still the certainty increased in his mind that there was some deceit, some fraud in connection with Sebastian Ritherdon's possession of Desolada and all that Desolada represented in value-- he did not hesitate now. As once he, with some of his bluejackets, had tracked slavers from the sea for miles inland and into the coast swamps and fever-haunted interior of the great Black Continent, so now he would track this man's devious and doubtful existence, as, remembering George Ritherdon's story, it seemed to him to be. If he had wronged Sebastian, if he had formed a false estimate of his possession of this place and of his right to the name he bore, no harm would be done. For then he would go away from Honduras for ever, leaving the man in peaceable possession of all that was rightly his. But, if his suspicions were not wrong---- He let himself down to the floor from the chair on which he had been sitting in the dark for now nearly an hour, and, quietly, noiselessly, he progressed along that solid floor--one so well laid in the past that no board either creaked or made any noise--and thus
- 87. he reached the balcony, there interposing nothing now between him and it but the lowered blind. Then when he had arrived there, he heard their voices plainly; heard every word that fell from their lips--the soft murmur of the woman's tones, the deeper, more guttural notes of the man. Only--he might as well have been a mile away from where they sat, he might as well have been stone deaf as able to thus easily overhear those words. For Sebastian and his companion were speaking in a tongue that was unknown to him; a tongue that, in spite of the Spanish surroundings and influences which still linger in all places forming parts of Central America, was not Spanish. Of this language he, like most sailors, knew something; therefore he was aware that it was not that, as well as he was aware that it was not French. Perhaps 'twas Maya, which he had been told in Belize was the native jargon, or Carib, which was spoken along the coast. And almost, as he recognised how he was baffled, could he have laughed bitterly at himself. What a fool I must have been, he thought, to suppose that if they had any confidences to make to each other, any secrets to talk over in which I was concerned they would discuss them in a language I should be likely to understand. But there are some words, especially those which express names, which cannot be translated into a foreign tongue. Among such, Ritherdon would be one. Julian, too, is another, with only the addition of the letter o at the end in Spanish (and perhaps also in Maya or Carib), and George, which, though spelt Jorge, has, in speaking, nearly the same pronunciation. And these names met his ear as did others: Inglaterra--the name of the woman Isobel Leigh, whom Julian believed to have been his mother, but whom Sebastian asserted to have been his; also the name of that fair American city
- 88. lying to the north of them--New Orleans--it being referred to, of course, in the Spanish tongue. So, he thought to himself, it is of me they are talking. Of me-- which would not, perhaps, be strange, since a guest so suddenly received into the house and having the name of Ritherdon might well furnish food for conversation. But, when coupled with George Ritherdon, with New Orleans, above all with the name of Isobel Leigh---- Even as that name was in his mind, he heard it again mentioned below by the woman--Madame Carmaux. Mentioned, too, in conjunction with and followed by a light, subdued laugh; a laugh in which his acuteness could hear an undercurrent of bitterness-- perhaps of derision. And she was this woman's relative, he thought, her relative! Yet now she is jeered at, spoken scornfully of by---- In amazement he paused, even while his reflections arrived at this stage. In front of where his eyes were, low down to the floor of the balcony, something dark and sombre passed, then returned and stopped before him, blotting from his eyes all that lay in front of them--the tops of the palms, the woods beyond the garden, the dark sea beyond that. Like a pall it rested before his vision, obscuring, blurring everything. And, a moment later, he recognised that it was a woman's dress which thus impeded his view, while, as he did so, he heard some five feet above him a light click made by one of the slats. Then, with an upward glance of his eyes, that glance being aided by a noiseless turn of his head, he saw that a finger was holding back the lath, and knew--felt sure--that into the darkness of the room two other eyes were gazing.
- 89. Welcome to our website – the perfect destination for book lovers and knowledge seekers. We believe that every book holds a new world, offering opportunities for learning, discovery, and personal growth. That’s why we are dedicated to bringing you a diverse collection of books, ranging from classic literature and specialized publications to self-development guides and children's books. More than just a book-buying platform, we strive to be a bridge connecting you with timeless cultural and intellectual values. With an elegant, user-friendly interface and a smart search system, you can quickly find the books that best suit your interests. Additionally, our special promotions and home delivery services help you save time and fully enjoy the joy of reading. Join us on a journey of knowledge exploration, passion nurturing, and personal growth every day! ebookbell.com