You are on page 1of 7

The Report is Generated by DrillBit Plagiarism Detection Software

Submission Information

Author Name ,Sreelakshmi


Title STUDY OF DOUBLE JPEG COMPRESSION DETECTION USIN..
Paper/Submission ID 669535
Submission Date 2022-12-12 16:35:50
Total Pages 4
Document type Research Paper

Result Information

Similarity 24 %
1 10 20 30 40 50 60 70 80 90

Sources Type Report Content


Words <
Internet
14,
1.26% Quotes
0.71%
3.87%

Journal/ Ref/Bib
Publicatio 12.24%
n 22.74%

Exclude Information

Quotes Excluded
References/Bibliography Excluded
Sources: Less than 14 Words Similarity Not Excluded
Excluded Source 0%
Excluded Phrases Not Excluded

A Unique QR Code use to View/Download/Share Pdf File


DrillBit Similarity Report

A-Satisfactory (0-10%)
B-Upgrade (11-40%)

24 8 B C-Poor (41-60%)
D-Unacceptable (61-100%)
SIMILARITY % MATCHED SOURCES GRADE

LOCATION MATCHED DOMAIN % SOURCE TYPE

1 Published by International Journal of Computer Science and Information Publication


11
Tech.- www.ijcsit.com

2 A bibliography on blind methods for identifying image forgery by Baba- Publication


2
2010

3 IEEE 2012 Fourth International Symposium on Information Science and Publication


2
by

4 Bibliography of digital image anti-forensics and anti-anti-forensics Publication


3
techniques by Qureshi-2019

5 IEEE 2012 IEEE International Symposium on Multimedia (ISM) - Publication


2
Irvine, by

6 moam.info Internet Data


1

7 Exposing Digital Image Forgeries by Detecting Contextual Abnormality Publication


1
Using Convo by Jang-2020

9 Detection of Double Compression in MPEG-4 Videos Based on Block Publication


1
Artifact Measure by He-2017
4
STUDY OF DOUBLE JPEG COMPRESSION  
DETECTION USING MACHINE LEARNING  
Lakshmi Santhosh V K1,Sreelakshmi K1, Aravinda Prabhu S1  
1 Department of IT,AIMIT, Mangalore, Karnataka, India  
lakshmimenon312@gmail.com,sreelakshmikizhakkool123@gmail.com, aravind@staloysius.ac.in  

ABSTRACT  
1
Double JPEG compression indicates that an image was originally a JPEG image and has been compressed by  
JPEG once again, which is common nowadays. We present an efficient machine learning-based method for  
differentiating between double and single JPEG compressed images. We employ several algorithms, including  
the Markov random process and the thresholding technique. The Markov random process is then used to model  
difference Two-Dimensional arrays in order to exploit second-order statistics. A thresholding technique is also  
used to decrease the size of the transition probability matrices that characterise the Markov random processes.  
7
All elements of these matrices are gathered as features for detecting double JPEG compression. This method can  
also be used to detect forgery parts in a double-compressed image, which is useful for blind authentication. The  
experimental outcomes display that the method can precisely locate the doctored part, and the experiment also  
shows that the higher the second compression quality is, the better the method detects the tampered areas. The  
detection efficiency is precise.  
Keywords: Double JPEG compression, Image, Matrices, Detection, Thresholding, Transition  

1.INTRODUCTION  
Double JPEG compression denotes that an image was originally a JPEG image and has been compressed by  
JPEG once again, as is common nowadays. The second jpeg compressed image has a different quality factor  
1
than the first. This is mainly used for digital forensics. Double JPEG compression tells the story of an image. In  
the literature, some methods for detecting double JPEG compression have been reported. In this study, we  
identify double JPEG compressed images using an effective scheme based on a machine learning framework.  
To generate features, the JPEG coefficient Two-Dimensional array of a given JPEG image is used here.  
Difference JPEG Two-Dimensional arrays, which are the magnitude differences between the JPEG coefficient  
Two-Dimensional array and its shifted versions in various directions. This is applied to extrude double JPEG  
compression artefacts. To take advantage of the second-order statistics, the Markov random process is then used  
to design these difference arrays. To significantly reduce the size of each transition probability matrix that  
characterises the Markov process, a thresholding technique is used. These matrix components are gathered as  
attributes for double JPEG compression detection. The support vector machine is utilised as a classifier.  

2.OBJECTIVE  
6
 Double jpeg compression detection using machine learning used to decrease the size of image without  
affecting the quality of image.  
1
 Detecting double JPEG compression is important in digital forensics.  
 Double JPEG compression tells the story of an image.  
 Detecting forgery parts in a double compressed image is critical work for blind authentication.  
3.METHODOLOGY  
Popescu's method and its effectiveness  
It is designed to detect double JPEG compression. Assume a JPEG mode was double quantized as part of the  
double JPEG compression operation. The letter b represents the first quantization step, and the letter a represents  
the second. If a/b is not an integer, double quantization introduces periodic artefacts into the JPEG mode's  
histogram.  
Feature creation  
A colour JPEG image has three components: Y, C b, and Cr. For the sake of simplicity, we only apply our  
feature generation scheme to the Y component in this paper. Each component, as shown in Figure, can be used  
to generate a subset of features using the similar procedure.  

JPEG 2-D array difference  


Observing the variation between an element and its neighbors in a JPEG Two-Dimensional array  
will improve double JPEG compression artefacts. F(u, v) (u[0, S h-1], v[0, S v-1]) denotes the  
Two-Dimensional array generated from a certain JPEG image, where S h is horizontal array size  
and S v is vertical array size. We can make four different Two-Dimensional arrays as follows:  

F h(u, v) = horizontal direction array  


F v(u, v) =vertical direction array  
F d(u, v)=main diagonal direction array  
F m(u, v)=minor diagonal direction array  

Matrix of transition probability  


We use a Markov random process to model the difference arrays defined above. A transition  
probability matrix can be used to characterize a Markov process, based on random process theory.  
We are using a thresholding technique to further reduce costs related to computation. Based on the  
statistical analysis of a large natural image dataset, we chose 4 as the threshold in our study. The one-  
step transition probability matrix elements associated with horizontal difference JPEG 2-D arrays are  
as follows:  

4. LITERATURE REVIEW/ RELATED WORKS  


2
Double-Compression Detection in JPEG2000 Images: In this paper, the statistical difference in the sub-  
band discrete wavelet transform coefficient histograms between single and double JPEG 2000 compression  
1
is examined, and a scheme for distinguishing between them is devised. Experiments demonstrate that the  
proposed method can detect double JPEG 2000 compression effectively and accurately.  
9 2 7
Review of Double Compression Detection Based on Markov Model of the First Digits of DCT Coefficients: In  
5
this paper, we propose using a Markov transition probability matrix to model the mode-based first digits of the  
Discrete Cosine Transform coefficients distribution and using its stationary distribution as features for double  
compression detection. The experimental shows the efficacy of this method, and a comparison has been made to  
show the improvement by using this method. A set of useful features for double compression identification is  
introduced in this study.  
An Improved Double Compression Method for JPEG Image Forensics: This paper present an improved double  
quantization detection method for detecting tampering in JPEG images.  
4 3
An Anti-forensic Algorithm of JPEG Double Compression Based Forgery Detection: A novel anti-forensic  
algorithm is used to a forensic analysis that can detect double JPEG compression with the similar quantization  
matrix.  

5.RESULTS  
We are using a large number of uncompressed images. The image which is uncompressed is first compressed to  
JPEG images using different quality factors it is Q1 in the Table l. The result is single JPEG compressed image.  
And each single JPEG compressed image is again compressed using a different quality factor it is Q2 in the  
Table l and the result is double JPEG compressed image. Q1 and Q2 are the quality factors. Table 2 indicates  
detection of accuracy.  
6.CONCLUSION  
In this study we presented and demonstrated the usefulness of our developed double JPEG compression  
detection scheme. Our paper is summarised as follows:  
 The first method is Popescu's method it is related on the "periodicity" pattern of histograms of JPEG  
4
modes along with double JPEG compression .It is designed to detect double JPEG compression  
 Differential JPEG 2D arrays in opposite directions increase rounding the errors, which are artefacts of  
double JPEG compression.  
 The Markov random process and its related transition probability matrix efficiently capture the  
statistical artefacts.  
 The thresholding method decreases the cost related to computation.  

REFERENCES  
[1] Y. Q. Shi, C. Chen, W. Chen, and M. P. Kaundiny, "Effect of recompression on attacking JPEG  
steganographic schemes ─ an experimental study", ISCAS07, New Orleans, LA, USA, May 27-30, 2007.  
[2] A. C. Popescu, "Statistical tools for digital image forensics", PhD thesis, Dartmouth College, Hanover, NH,  
USA, December 2004 (advised by H. Farid). http://www. cs.dartmouth.edu/farid/publications/apthesis05.pdf.  
[3] J. Lukas and J. Fridrich, "Estimation of primary quantization matrix in double compressed JPEG images",  
Digital Forensic Research Workshop, Cleveland, OH, USA, August 2003.  
[4] D. Fu, Y. Q. Shi, and W. Su, "A generalized Benford's law for JPEG coefficients and its applications in  
image forensics", SPIE Electronic Imaging, San Jose, CA, USA, 2007.  
[5] A. Leon-Garcia, Probability and random processes for electrical engineering, 2nd Edition, Addison-Wesley  
Publishing Company, 1994.  
[6] C. C. Chang and C. J. Lin, "LIBSVM: a library for support vector machines", 2001. Available online at  
http://www.csie.ntu.edu.tw/~cjlin/libsvm.  
[7] http://vision.cs.aston.ac.uk/datasets/ucid/ucid.html.  
[8] http://ist.sysu.edu.cn/en/index_en.asp.  
[9] http://photogallery.nrcs.usda.gov/.  

You might also like