Secure Video Processing using ROI Extraction and ECC Encryption

Available online at https://www.ijasrd.org/
International Journal of Advanced Scientific
Research & Development
Vol. 06, Iss. 03, Ver. I, Mar’ 2019, pp. 24 – 33
Cite this article as: Aravindan, S., Pavithra, R., Sindhuja, J., & Sobika, R., “Secure Video Processing using ROI Extraction
and ECC Encryption”. International Journal of Advanced Scientific Research & Development (IJASRD), 06 (03/I), 2019, pp. 24
– 33. https://doi.org/10.26836/ijasrd/2019/v6/i3/60304.
* Corresponding Author: S. Aravindan, aravindan@egspec.org
e-ISSN: 2395-6089
p-ISSN: 2394-8906
SECURE VIDEO PROCESSING USING ROI EXTRACTION AND
ECC ENCRYPTION
S. Aravindan1*, R. Pavithra2, J. Sindhuja2 and R. Sobika2
1 Asst. Prof., Department of Computer Science and Engineering, E.G.S. Pillay Engineering College,
Nagapattinam.
2 Final Year, Department of Computer Science and Engineering, E.G.S. Pillay Engineering College,
Nagapattinam.
ARTICLE INFO
Article History:
Received: 18 Mar 2019;
Received in revised form:
29 Mar 2019;
Accepted: 29 Mar 2019;
Published online: 10 Apr 2019.
Key words:
Surveillance Video,
ROI Encryption,
Layered Cellular Automata,
Reversible Rule,
Shift Transformation.
ABSTRACT
Video encryption is the process of encrypting videos to hide
the object in video for secure video sharing. To hide privacy
from sensitive areas in video frame, ROI extraction
techniques are proposed. Lightweight encryption algorithm is
performed on every ROI to make the privacy sensitive
information. To implement ECC encryption for encryption of
extracted ROI. Provide secure video communication between
sender and receiver.
Copyright © 2019 IJASRD. This is an open access article distributed under the Creative Common Attribution
License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original
work is properly cited.
INTRODUCTION
Network security is the security provided to a network from unauthorized access and
risks. It is the duty of network administrators to adopt preventive measures to protect their
networks from potential security threats.
With the rapid development of network and communication technology, the
application of surveillance systems have significantly increased in the last two decades.
Surveillance cameras are deployed not only in public places such as airport, bus station, but
also in private places such as home, office, to name a few. By surveillance videos, the
authenticated users can identify most of objects they want to observe or track with. For
example, the users can see the live or stored surveillance videos on hand-held devices to
know what is happening or has happened in their house. However, as it’s possible to obtain
the photos of humans’ faces, the plate numbers of the cars among the monitoring area, the
privacy information is then jeopardized through the surveillance cameras in public areas.
The private places like homes are vulnerable to the surveillance videos too. Generally,

Aravindan et al., (2019)
25 Volume 06, Issue 03, Version I, Mar’ 2019
encryption is an effective method to protect privacy as well as security. To hide privacy from
sensitive areas, different cryptographic and ROI extraction techniques are presented. As a
result, hardly is it integrated with an IOT device at the surveillance cameras’ side.
Although video are made from frames, where each frame is an image with a
compressed format, some inherent features such as redundancy, bulk data capacity and
high correlation between pixels make the typical image cryptosystems unsuitable for videos.
PRELIMINARIES
A CA is a dynamic data handling framework whose space, time and state are for the
most part discrete. The CA comprises of a few cells organized in a customary matrix. Every
cell has its very own state and all cells refresh their states synchronously as per a
presetting nearby principle. The new condition of a cell depends without anyone else state
ever, yet additionally the conditions of its neighbors'. For a limited CA, the occasional limit
conditions are generally connected, where the limit cells are linked as a falling framework
and consequently the CA can be treated as a limited state machine.
For an1D CA, its nearby progress rule f is characterized as pursues:
Where st I indicates the condition of ith cell at time t and r is neighborhood range
with the inside as st I. In the event that there are two cell states just, absolutely f has 22r+1
diverse data sources pursued by 222r+1 distinctive neighborhood change rules.
Basic CA (ECA) is the least difficult CA that every cell just has two conceivable
states and three neighbors. There are 28 = 256 basic standards altogether and every one of
them is listed by a whole number untruths in [0,255]. All conceivable info conditions of f are
organized in the request as 111,110,••• ,001,000, where the subsequent yield states framed
a twofold succession and the relating rule number is spoken to by the picked decimal
numbers. The model basic tenets are appeared in Table 1.
Table – 1: Elementary Rule 15,30 and 90
Rule 111 110 101 100 011 010 001 000
15 0 0 0 0 1 1 1 1
30 0 0 0 1 1 1 1 0
90 0 1 0 1 1 0 1 0
Reversible Cellular Automata (RCA) is a special case of CA whose transition rule is
reversible and each state has only one successor and one predecessor. RCA is especially
suitable for cryptosystem because the reversibility property of CA ensures that any
encrypted message can be decrypted by the same algorithm performed in the reverse
direction. It has been proved that whether a rule of an 1D CA is reversible is determined,
whereas for a two or more dimension CA the same reversibility is un decidable. A layered
cellular automata (LCA) is a cascading system of 2D CAs with the same size, where each
layer is a composition of 1D CAs. The structure of LCAs makes a more complex but flexible
expression ability.
In this paper, an 8-layer CA is devised by us to encrypt the RoIs of the gray
surveillance video frames because the 8-layer CA can match up with the binary pixel matrix
of a video frame completely.

Secure Video Processing using ROI Extraction and ECC Encryption
Volume 06, Issue 03, Version I, Mar’ 2019 26
THE FRAMEWORK OF OUR APPROACH
Our approach is a hierarchical privacy protection method for surveillance videos now
can apply to the video format H.264. By our approach, the non-authenticated users can only
see the real-time surveillance video without ROIs, while the authenticated users can watch
the complete surveillance video through an on-demand manner.
The ROIs containing privacy sensitive information such as faces and license plates
are extracted from the I frames (JPEG format) through the ROI detection and segmentation
methods in. It’s possible that there are more than one ROI in a video frame. The length and
width of each ROI is the multiples of 16, so that each ROI can be divided into the sub-blocks
of the size 16×16.
LCA based lightweight encryption algorithm is performed on every ROI to make the
privacy sensitive information unrecognizable, and then all ROIs from a frame are encrypted
independently. Hence, if there are some bits spoiled during the encoding and/or
transmission, it will only affect the decryption within their blocks, which limits the errors
propagation. All ROI blocks in a video frame are encrypted by the same random keys, which
can reduce the consumption of the key’s exchange and storage. Note that the ROIs from
different video frames are required to use different encryption keys. After the ROIs are
extracted, the pixel values with all (0 or 1) are filled back to replace the extracted ROI
regions in the I frames. In what follow, the new surveillance video based on the new I
frames is then re-encoded and transmitted, which can be watched real-time on a computer
or a handheld device with network service. Whereas the encrypted ROIs are stored at the
camera side without compression and be used for the on-demand service later, i.e., the
stored ROI will be transmitted, decrypted and then integrated with the new I frames to
recover the original surveillance video if and only if the authenticated users require to do so.
Note that the encrypted ROI data lost almost all spatial correlations, compression cannot
decrease the data size remarkably. The ROIs encryption and the new video’s encoding are
carried out independently, which make the proposed method can not only protect privacy
information, but also meet the real-time requirement. In addition, since the ROIs in the
original I frames are replaced with regions that pixel values are all 0 or 1, the surveillance
video data compression ratio is increased because the correlations of these data are
increased. The cost of our approach is the affordable delay for the authenticated users to
retrieve the complete surveillance video.
ECC AND ROI ENCRYPTION ALGORITHM
 Propose a lightweight LCA-based method to encrypt the privacy sensitive ROIs for
surveillance videos.
 It allows extending the existing surveillance systems without modifying the camera's
hardware.
 It focuses on the ROI privacy protection with lightweight cryptographic techniques.
 Input video is convert into frames.
 To extract the ROIs from video frames is not highlighted.
 A ECC-based approach is devised to encrypt ROIs for privacy protection in
surveillance videos.

(A) Key Generation
We first pick four sets of reversible ECA rules,{15,85}, {51,51}, {170,240}, and
{204,204}, as the change guidelines, and afterward such principles are filed by a mapping
capacity as following:
F : {00, 01, 10, 11}→{15/85, 51/51, 170/240, 204/204}
Where F is an injective function. Furthermore, in each pair of rules, the one is
utilized for encryption and the other is utilized for decoding. A pseudo arbitrary double
succession PR is created from a seed number, which is secretly shared between the
encryption and the decryption sides. The PR is divided into blocks of two bits, where each
square speaks to a specific guideline for a 1D CA in the LCA. In this way, the length of the
succession PR is controlled by the quantity of 1D CA in the LCA, where the number is
decided by the height of its ROI block .specifically , the measure of a ROI square is 16×16,
each layer of the layered CA has 16 lines and after that the length of the arrangement PR is
8×2×16=256. The double succession PR with the cycle number N will be utilized as the
encryption and decoding key.
(B) Encryption
The ROI squares are scrambled as pursues.
Stage 1: Initialization. Arrange the binary sequence from the original RoI block
into an 8-layerCA , and then let each layer contain 16×16 bits.
Stage 2: Rule advancement. Each layer is treated as a composition of the row shift
the 1DCA switch the same size. There after, cells from different 1DCAs change their states
independently according to different ECA rules which selected by the blocks in the
succession PR.
Stage 3: Intra-layer move. A half move change is performed in each layer. For
each line in a layer, just half cells change their states to the conditions of the cells at their
adjoining line. specifically, the phones at back segments will change their states and
alternate cells will keep static. In the event that there are k(k ∈ N+) segments in each layer,
the condition of it
If there are 𝑘 𝑘 ∈ 𝑁+ columns in each layer,the state of 𝑖 𝑡ℎ row in the 𝑙 𝑡ℎ layer at time 𝑡 is 𝑆𝑙,𝑖,1,⋯,
𝑡
𝑆𝑙,𝑖,𝑗,⋯,
𝑡
𝑆𝑙,𝑖,𝑘
𝑡
.
It will be shifted to the new state 𝑆𝑙,𝑖,1,⋯,
𝑡+1
𝑡+1
𝑆𝑙,𝑖,𝑘
𝑡+1
after an intra − layer half shift transformation,where
Note that such a change makes the half cells in each layer change their states , and
it indirectly inﬂuences the cell snot shifted because such transformation can change their

neighbors' states. In this manner, if a cell in succession changes its state, it will influence
the cells in indistinguishable column from well as in alternate lines.
Stage 4: Inter-layer move. It is a half move change between adjoining layers.
Furthermore, this is an occasional change, where half cells in a single layer will move to its
upper layer and the cells in the first layer will move to the last layer. Specifically, when the
between layer half move change has occurred, the new conditions of the ith push in the lth
layer is 𝑆𝑙,𝑖,1,⋯,
𝑡+1
𝑡+1
𝑆𝑙,𝑖,𝑘
𝑡+1
at time t +1, where
𝑆𝑙,𝑖,𝑗
𝑡+1
=
𝑆𝑙+1,𝑖,𝑗,
𝑡
1 ≤ 𝑗 ≤
𝑘
2
𝑆𝑙,𝑖,𝑗,
𝑡 𝑘
2
+ 1 ≤ 𝑗 ≤ 𝑘
Stage 5: Intra-push irregular move. Each 1D CA in a layer performs a periodic
intra-row shift transformation. All cells in a1DCA shift to their left according to a random
shift number. Give stomach muscle a chance to indicate the progress standard of a 1D CA
selected by the parallel grouping in the key, where a,b ∈ {0,1}, at that point the move
number of the CA is defined as a × 21 + b × 20 = 2a + b. Thusly, each 1D CA in the layered
CA has four conceivable move numbers. Rehash Step 2 to Step 5 until the pre-defined cycle
number came to. At long last, the scrambled ROI square is worked by changing over the
final condition of the 8-layer CA into a pixel framework. Fig. 3 demonstrates the previously
mentioned encryption process. Calculation 1 is utilized for ROIs encryption.
Table – 2: Move Numbers with its Comparing Rule Lists
Rule Index 11 10 00 01 00 10 11 10
Shift Number 3 2 0 1 0 2 3 2
Figure – 3: The LCA-based Surveillance Video Encryption Algorithm
Encryption Algorithm:-
 Suppose sender wants to send a message m to the receiver

o Step 1: Let m has any point M on the elliptic curve
o Step 2: The sender selects a random number k from [1,n-1]
o Step 3: The cipher texts generated will be the pair of points (B1,B2) where
B1= k*G B2= M + (k*G)
(C) Decryption
The decoding key is equivalent to the encryption key. The advancement rules spoken
to by the arrangement are the turn around principles of the ones utilized in the encryption.
The binary sequence of an encrypted ROI block is set as the starting condition of a 8-layer
CA. Forward moving uses the principles chosen by the decoding key. From there on, the
converse changes of the three unique changes referenced above are performed, as well.
Figure – 4: An Example of Random Shift
1 0 0 0 0 1 1 0

0 0 1 1 0 1 0 0
1 0 0 1 1 0 1 0 0 1 1 0 1 0 1 0
1 1 0 1 1 0 0 0 1 1 0 1 1 0 0 0
0 1 1 0 1 1 0 0 1 1 0 1 1 0 0 0
0 1 0 0 0 1 0 1 0 1 0 0 0 1 0 1
0 1 1 0 1 1 0 0 1 0 1 1 0 0 0 1
0 1 0 0 1 1 1 1 0 1 1 1 1 0 1 0
1 0 0 0 1 1 0 1 0 0 1 1 0 1 1 0
(a) (b)
Decryption Algorithm
 To decrypt the cipher text, following steps are performed:-
o Step 1: The receiver computes the product of B1 and its private key
o Step 2: Then the receiver subtracts this product from the second point B2 M
= B2- (dB * B1) M is the original data sent by the sender.
ARCHITECTURE

LIST OF MODULES
(a) Video Processing
(b) ROI Extraction
(c) ROI Encryption
(d) Video Retrieval
(a) Video Processing
 Videos are made from frames, where each frame is an image with a compressed
format;.
 As each video consisting of its own number of frames decided by the size of the video,
whenever converting the video to frames, the frames requires more memory than
that of the original video.
 The surveillance video is composed of a sequence of group of pictures.
(b) ROI Extraction
 The ROIs containing privacy sensitive information such as faces are extracted
through the ROI detection and segmentation methods.
 It's possible that there are more than one ROI in a video frame.
 LCA-based lightweight encryption algorithm is performed on every ROI to make the
privacy sensitive information unrecognizable.,
 ROIs from a frame are encrypted independently.
(c) ROI Encryption
 Extracted ROI parts are encrypted with help of ECC encryption algorithm.
 Whereas the encrypted ROIs are stored at the server side.
 The stored ROI will be transmitted, decrypted to recover the original surveillance
video.
 The ROIs encryption and the new video's encoding are carried out independently,
which make the proposed method can not only protect privacy information, but also
meet the real-time requirement.
(d) Video Retrieval
 The decryption key is the same as the encryption key.
 The evolution rules represented by the sequence are the reverse rules of the ones
used in the encryption.
 Thereafter, the inverse transformation was performed.
SECURITY AND PERFORMANCE ANALYSIS
Encryption calculation utilized for security assurance in observation video ought to
be sufficiently secure to avoid assailants getting the protection data from the scrambled
recordings. In our work, locales contain private data in edges are scrambled by our LCA-
based methodology and the non-security areas will stay flawless. A few factual tests and
estimations are utilized to break down the security of our methodology.

(A) Data Entropy
Data entropy is an essential estimation for the dimensions dispersion in a picture.
The entropy of a picture I is defined as pursues:
𝐻 𝐼 = − 𝑝 𝑥𝑖 log2 𝑝 𝑥𝑖
𝐿
𝑖
Where L is the quantity of dark dimensions, xi is the ith dim an incentive in the
picture I, p(xi) is the event likelihood of xi, and pi p(xi)=1.
The higher the entropy is,the more the dark dimensions near uniform circulation
are. For a picture satisfies uniform conveyance, there are 256 dark dimensions with a
similar event probability, accordingly the ideal entropy esteem is 8 and the entropy of an
encoded picture ought to be way to deal with such an esteem. Table 3 demonstrates a
correlation of the entropies of three dark pictures and their relating figure pictures encoded
by our LCA-based technique and some l-------different strategies. In Table 3, the entropies of
the pictures scrambled by our strategy are near the ideal esteem 8, which exhibits that our
technique is superior to that of alternate strategies.
(B) Histogram
A picture histogram straightforwardly displays the factual qualities of the picture
pixel values. Generally, the pixel estimations of the encoded pictures need to fulfill a
uniform irregular conveyance.
(C) Relationship of Adjacent Pixels
Correlation trial of the neighboring pixels is an imperative factual technique to
assess the dispersion and disarray of an encryption calculation. Scrambled pictures ought to
have a just about zero connection between's the neighboring pixels, though the plain
pictures should display a solid relationship. To play out a connection test on a picture, we
haphazardly pick 1,000 pairs of adjacent pixels in vertical, diagonal and horizontal
directions from a plain image and its cipher image, separately. We at that point compute
the relationship coefficient, where the connection coefficient is defined as pursues:
𝑟𝑥𝑦 =
𝑐𝑜𝑣 𝑥, 𝑦
𝐷 𝑥 . 𝐷 𝑦
𝐷 𝑥 =
1
𝑛
𝑥𝑖 − 𝐸 𝑥
2
𝑛
𝑖=1
𝐸 𝑥 =
1
𝑛
𝑥𝑖
𝑛
𝑖=1
𝑐𝑜𝑣 𝑥, 𝑦 =
1
𝑛
(𝑥𝑖 − 𝐸 𝑥 )(𝑦𝑖 − 𝐸 𝑦 )
𝑛
𝑖=1
(D) Differential Analysis
A differential analysis refers to the influence of slight change in plain picture
concerning the figure picture. In the event that a solitary pixel's change can have a
significant influence in the figure picture, that the encryption calculation is secure against
the differential assault is finished up. Two lists are utilized to quantify such an influence in

the paper, viz., the number of pixels change rate (NPCR) and the unified normal evolving
power (UACI), where:
𝑁𝑃𝐶𝑅 =
𝑖,𝑗 𝐷 𝑖, 𝑗
𝑀 × 𝑁
× 100%
𝑈𝐴𝐶𝐼 =
1
𝑀 × 𝑁
𝐶1 𝑖, 𝑗 − 𝐶2 𝑖, 𝑗
255
𝑖,𝑗
× 100%
𝐷 𝑖, 𝑗 =
0, 𝐶1 𝑖, 𝑗 = 𝐶2 𝑖, 𝑗
1, 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
(E) Key Sensitivity
As indicated by Shannon's hypothesis, encryption result should be delicate to any
adjustments in a key. In our calculation, the key is made up of a pseudo random binary
sequence and an iteration number. To test the key's affectability, the accompanying
analyses are performedona 64×64 bits grayimage 'rice', viz., Fig.10(a). We first encoded the
picture with K0(PR,N = 5) to get a reference figure picture C0, where PR is a 16 × 64 bits
twofold succession, and after that we encode the plain picture with two modified keys
K1(PR1,N = 5) and K2(PR,N = 7), where the contrast among PR and PR1 is just a single
piece. The comparing figure pictures are signified by C1 and C2 respectively, as shown in
Fig.10. In order to present a quantitative illustration, correlation co-efficients between C0
and C1, C0 and C2 are calculated. The values are −0.0196 and 0.0021, respectively, which
show that our LCA-based approach is key delicate.
(F) Animal Force Attack
Animal power assault or thorough assault is an essential system of attempting each
conceivable key thusly until the right key is identified. Hypothetically, the extent of the key
space decides the down to earth practicality of playing out a savage power assault.
Note that a casing may has more than one RoIs. So as to lessen the expense of the
key's circulation, all RoI obstructs in a casing will be encoded with a similar key. Since our
methodology satisfies the necessities of perplexity and dispersion, regardless of whether two
squares which have minor contrasts are encoded by a similar key, the outcomes are far
various. Further more, a RoI extracted from a frame may be the same as the one from the
other frames, those frames will be encrypted by various keys and thus even similar RoI will
demonstrate diverse encryption results. Accordingly, it's computationally infeasible to
dispatch a key beast drive assault to our methodology.
(G) Execution Analysis
Since CA are locally parallel, CA-based encryption plot is very efficient in both
equipment and programming executions. In the lightweight LCA-based iterative picture
encryption technique proposed, every cell performs rule evolution, transposition and XOR
operations for N cycles, the computational multifaceted nature is O(N). The aftereffects of
the execution examination demonstrated that it is more efficient when contrasted and some
1D and 2D based encryption strategies. In our LCA-based method, fewer operations are
performed for every cell in all RoI hinders than the one. Just a single guideline
advancement and three move changes are engaged with each iteration. For N iterations, the
computational complexity of our technique is O(N), as well. Since the RoIs in each casing
are composed into squares and be encoded freely and synchronously, which add to the

efﬁciency of our approach. Further more, the results of the above analysis show that even
less cycle rounds empower the scrambled squares to have a decent factual trademark.
TEST RESULTS
The investigations dependent on the observation recordings of H.264, as showe din
Fig.5(a) and (b), where the RoIs contain privacy delicate data, for example, human face and
vehicle plate number. In our tests, the RoIs are removed by the strategy. The proposed
methodology is connected on these extricated RoIs, the encoded RoIs are appeared in Fig.
5(c) and (d), the non privacy locales are then incorporated with the substitution districts
(white area) for security assurance are appeared in Fig. 5 (e) and (f) and the recouped total
video outlines for the verified clients are appeared in Fig. 5 (g) and (h).
CONCLUSION
Proposed approach of video encryption provides more secure communication and
video transaction between sender and receiver. The proposed method satisfies the real-time
requirements of secure data and videos sharing.
REFERENCES
[1] Auer, S., Bliem, A., Engel, D., Uhl, A., and Unterweger, A., (2013) “Bitstream-based
JPEG Encryption in Real-time”. International Journal of Digital Crime and
Forensics (IJDCF), 5 (3), pp. 1 – 14.
[2] Kanso, A., and Ghebleh, M., (2012) “A novel image encryption algorithm based on a 3D
chaotic map”. Communications in Nonlinear Science and Numerical Simulation, 17
(7), pp. 2943 – 2959. DOI: https://doi.org/10.1016/j.cnsns.2011.11.030.
[3] Dufaux, F., (2011) “Video scrambling for privacy protection in video surveillance: recent
results and validation framework”. Proceedings of SPIE - The International Society
for Optical Engineering 8063. DOI: https://doi.org/10.1117/12.883948.
[4] Tralic, D., and Grgic, S., (2016) “Robust Image Encryption Based on Balanced Cellular
Automaton and Pixel Separation”. Radioengineering, 25 (3), pp. 548 – 555.
[5] Chai, X., Gan, Z., Yuan, K., Chen, Y., and Liu, X., (2017) “A novel image encryption
scheme based on DNA sequence operations and chaotic systems”. Neural Computing
and Applications, 31(1), pp. 219–237. DOI: https://doi.org/10.1007/s00521-017-2993-
9.

Secure Video Processing using ROI Extraction and ECC Encryption

More Related Content

Similar to Secure Video Processing using ROI Extraction and ECC Encryption (20)

More from IJASRD Journal (20)

Recently uploaded (20)

Secure Video Processing using ROI Extraction and ECC Encryption