Digital Image Sham Detection Using Deep Learning

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 10 Issue: 04 | Apr 2023 www.irjet.net p-ISSN: 2395-0072
© 2023, IRJET | Impact Factor value: 8.226 | ISO 9001:2008 Certified Journal | Page 1816
Digital Image Sham Detection Using Deep Learning
Mr. Hemanth C1,Ms. Divya A Srivathsa2, Ms. Gouthami M3, Ms. Monica S4, Ms. Sarika B V5
1 Assistant Professor, Dept. of Computer Science and Engineering, Maharaja Institute of Technology,
Thandavapura
2,3,4,5 Students, Dept, of Computer Science and Engineering, Maharaja Institute of Technology, Thandavapura
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - “Digital Image Sham Detection Using Deep
Learning”, Capturing images day by day as been increasing
since there are availability of variety of cameras. Images as
become a part in our daily lives because they contain an lot
of information and sometimes it is also required to capture
extra images to find additional information. This increases
the grievousness and recurrence of fake image, which is now
a major source of concern. A lot of customary techniques
have been come into being over time to detect image
falsification. In recent years, convolutional neural networks
(CNNs) have come across much intentness, and CNN has also
supremacy the field of image forgery detection. Even so,
most image falsification techniques based on CNN that
survive in the literature are limited to detecting a distinct
type of sham . As a result, a technique capable of logically
and well aimed detecting the presence of out of sight
forgeries in an image is required.
Key Words: Image, Detection, CNN
1.INTRODUCTION
Now-a-days a handful of software are accessible that are
used to exploit image so that the image is a look alike of the
unedited. Images are cast-off as substantiate galley for any
offence and if these image does not remain veritable then it
will cause an issue. In this scientific era a large number of
people have become casualty of image falsification.
A large number of people operate technology to
modify images and use it as verification to mislead the
court. Image manipulation is any type of operation that is
accomplished on digital images by utilizing any software, it
is also mentioned as image polish. So, to end to this, all the
images that are allocated through social media should be
designated as original or fraud errorless.
Social media is a huge party line to mingle, split
and widen knowledge but if heedfulness is not employed, it
can misguide people and even cause devastation due to
unwitting false advocacy. Image tampering is a type of
image falsification which return some content of an image
with up to date content. If the up to date content is
emulated from the same image itself then it is called copy-
move tampering and if the up to date content is emulated
from non-identical image then it is known as image
splicing.
1.1 Overview
Numerous methods have been uplifted in the
literature to compact with image falsification. The large
number of conventional methodology are based on specific
artefact left by image falsification, whereas fresh
techniques based on CNNs and deep learning were
established, which are brought up below. First, we will
mention the various orthodox techniques and then
progress on to deep learning based techniques. It provides
two level inspection for the image. At first level, it examine
the image metadata. Image metadata is not that much
authentic since it can be changed using effortless
programs. But most of the images we come across will
have nonchanged metadata which helps to figure out the
changes.
1.2 Problem Statement
Since the innovation of photography, individuals
and company have often look for paths to modify and
manipulate images in order to defraud the viewer. Existing
systems have worked on the contrast of image falsification
identification methods, these are frequently narrowed in
span and only weigh up alternate of the identical algorithm
on images that are expressly fabricate for that type of
routine. There are also shamed images which cannot be
identified by the existing applications.
2. EXISTING SYSTEM
In existing forgery image detection system, it can be use
to detect only limited type of image forgery like splicing
and copy-move and not able to detect all types of forgery
images.
Using new technologies any images can be forged with
help of variety of tools available in the internet which
makes impossible for humans to differentiate whether an
image is forged or not.
Even with the help of complex neural network it is nearly
impossible to determine forged or not.

3.PROPOSED SYSTEM
In this proposed system the application is able to
detect whether an image is forged or not for all types of
forged images like copy-move, splicing, tampering,
morphing etc., here the application uses VGG16 and VGG19
algorithm and with the help of learning rate 0.0001 the
VGG16 and VGG 19 algorithm gives 100% accuracy and for
comparison purpose this system also uses error detection
analyses also. The process how application works are
-First it will train the model using the provided
datasets.
-Then in testing the user can choose one analyses
type out of three and as to put an image for the test then
the result is published in the form of Pie Chart.
The main advantage of this system is the user can clearly
compare with each algorithm to check the image
originality and then decide what to do with that image.
4. SYSTEM ARCHITECTURE
Fig-1 Architecture of Image Forgery
The system architecture defines the way how the
system is designed. It also defines the relationship with
other components and other aspects of software and
reflects how it interacts with other systems and outside
world. The architecture above describes the proposed
system. It describes the way this system is developed and
how it is connected to other components and the working
flow of it.
5. Need of Digital Image Forgery Detection
The image forgery detection is very important nowadays
because of rapid growth in the technology field there are
many tools using which any one can tamper the original
image and it will be very harmful if they use it in a bad
way. So, it is very important to distinguish between
authenticate image and the fake image which human
cannot do it with their eyes. Image forgery detection is
important in many aspects such as,
Maintaining Authenticity: Most of the images are often
used as evidence in legal and investigative contexts as well
as in journalistic and documentary contexts.
Preventing misinformation: In today’s generation any
image can be forged according to the needs and can spread
false information for the society which is very dangerous.
With the help of detection system this can be prevented.
Protecting Intellectual Property: Image forgery can also be
used to steal intellectual property of an artist. Detecting
image forgery can help protect the rights of the creator.
Overall, image forgery detection is essential for maintain
the integrity of images and ensuring that they are used
appropriately and accurately in a variety of contexts.
6. IMAGE FORGERY TYPES
The image may be forged either by adding,
removing or replacing some regions in the original image
with only one thing in mind that it leaves no visually
detectable trace. The image can be forged by using several
methods, these methods are commonly categorized
Fig-2 Types of Image Forgery
6.1 Copy Mover Forgery
Copy-Move Forgery means duplicating some part
of the image and replacing in other part of the same image
as shown in below figure. The intention of this is to conceal
some part of the image information. It is the most usually
utilized methods to forge an image. As the forged part of
the image remains in the same image itself. Therefore, its
detection is usually tough.
Fig-3 Effect of Copy Move Forgery and Image
Retouching

(a)First Image (b) Second Image (c) Forged Image
6.2 Image Forgery, using retouching
It is the process of combining more than one
image. The images are combined to make an altered image.
It uses cut/copy paste operations. A bit of one image is
taken and pasted onto some other image. In order to
completely connect the cut/copied part of an image into
another image as shown in the above figure, it need some
additional postprocessing operations. The pasted portion
alters the pattern of the image. Thus, analysis of image
pattern helps in detection of image forgery.
7. ALGORITHM
Fig-4 Convolution layers of VGG-16 Algorithm
Step 1:
Input Image: An image from the training datasets is taken.
Step 2:
Image Processing: Scale down the image pixel and convert
them into numpy.
1. Filtering: Suppress the high frequency and
smoothen the image.
2. Padding: To have zero padding so that the output
does not differ from the input image.
Step 3:
Data preprocessed: Flipping the images vertically and
horizontally.
1. 2D/3D convolution: To perform element wise
multiplication.
2. Pooling: (Ih-f+1) / S*Ic (Ih- Image height, Iw-
Image width, Ic- Number of channels in feature, f-filter,
sStride length)
Step 4:
Activation function: Based on the test cases it activates the
model along with background verification.
Step 5:
Output: Predicts the image score whether the image is
original image or forged.
Step 6:
End
8. MODULE DESCRIPTION
i. Tensor Flow
A free and comprehensive open-source software library
for artificial intelligence and machine learning is called
TensorFlow. The creation and training of machine learning
models use it.
ii. Keras
The Keras high-level Python library runs on top of the
TensorFlow framework and is small, simple to learn and
effective. It is made with an emphasis on comprehending
deep learning methodologies.
iii. pyQt
It is a python binding for Qt, a collection of libraries and
development tools that offer abstractions for graphical
user interfaces regardless of platform.
iv. Pillow [Pi]
All the fundamental image processing capabilities are
available in the pillow library. It supports a wide range of
picture file types for opening, editing and saving.
v. Epoch
The entire number of interactions of all the training data in
one cycle for training the machine learning model is
referred to as all the training data and is utilized all at
once.

9. FLOWCHART
Fig-5 Flowchart of Image Forgery Detection
10. EXPERIMENTAL RESULTS
In this project we are majorly using VGG-16 and
VGG-19 algorithm. The reason behind using these
algorithms is that the accuracy of these two algorithms is
too high in comparison to others. Also, this project delivers
98.8% accuracy to all the datasets provided. The results
are shown as below.
(a) (b)
(c) (d)
Fig-6 Snapshot Results of the Experiment
(a) Window of Image Forgery detection.
(b) Epoch for training datasets.
(c) Result of training data.
(d) Epoch for testing datasets.
(e) Result of testing datasets.
ACKNOWLEDGEMENT
(e)
11. CONCLUSIONS
In this study, multiple passive picture forgery
detection methods were sketched out. A thorough
examination of several forgery detection methods is also
provided. In addition, this publication offers a variety of
data sets that may be used with various forgery detection
strategies. The primary shortcoming of currently available
forgery detection methods is that they require human
intervention in order to be detected. The failure of the
procedures up to this point to distinguish between
malicious and lawful tampering is another significant flaw
in their design. Additionally, the discussed methods are
only designed to detect the forgery type for which they
were developed. Therefore, a comprehensive, reliable
method to spot any kind of image forgery is required A
potential solution for digital picture forensics is proposed
with the development of powerful artificial intelligence
algorithms. Although deep-learning-based methods show
promise, they lack the power to perform well in a number
of digital image forensics applications. All of these
parameters still require a lot of work to be done.
We would like to give Prof. Hemanth.C, our
project leader, our heartfelt appreciation for leading us
through this project, for sharing his invaluable insights
and recommendations with us, thus helping us better it
beyond our wildest expectations. Secondly, we would like
to express our gratitude to our project coordinators, Dr.
Ranjit K. N. and Dr. HK Chethan, who continuously
encouraged us and assisted us in completing this
project within the allotted time frame. We also want
to extend our sincere gratitude to our department's
head, Dr. Ranjit KN, for giving us a venue to work on

REFERENCES
projects and show how our academic curriculum is put
to use in the real world. We would like to thank Dr.
Y T Krishne Gowda, our principal, for giving us the
chance to complete this amazing project on the
topic of "Digital Image Sham Detection." This project has
also assisted us in conducting extensive research and
learning how to put it into practice.
[1]Mukesh M.Goswami and Zankhana J. Barad, [2020]
"Image Forgery Detection Using Deep Learning",
[2] P. Lavanya, B. Jagruti, M. Srinidhi, and P.
Mallesh, "Image Forgery Detection", [2022]
[3] A review of "A review on Digital Image
Forgery Detection" by Jahnavi Ega, Deepak Sri Sai
Krishna, and V.M. Manikandan was published in
[2021].
[4] Kshitij Swapnil Jain, Udit Amit Patel, and
Rushab Kheni, "Handwritten signatures forgery
detection", [2021]
[5] Francesco Marra, "A Full-Image resolution
End-toEnd Trainable CNN Framework for Image
Forgery Detection"[2020]
[6] "Detection of Image Forgery" [2020] by
Shubham Sharma and Sudeeksha Verma
[7] S Prayla Shyry, Mahitha Moganti, and
Saranya Meka, "Digital Image Forgery Detection"
[2019].
[8] "Copy-Move Forgery Classification via
Unsupervised Domain Adaptation" by Akash Kumar and
Arnav Bhavsar was published in 2019.

Digital Image Sham Detection Using Deep Learning

More Related Content

Similar to Digital Image Sham Detection Using Deep Learning (20)

More from IRJET Journal (20)

Recently uploaded (20)

Digital Image Sham Detection Using Deep Learning