Kahane - Revision history

imported>Projects221: /* Appendix 1 */

2013-03-20T16:43:33Z

Appendix 1

← Older revision		Revision as of 16:43, 20 March 2013
Line 161:		Line 161:

	Inconclusive code for detecting patterns in decompressed coefficients		Inconclusive code for detecting patterns in decompressed coefficients
	[[File:~~Testing~~.zip]]		[[File:Inconclusive_Test.m.zip]]

imported>Projects221: /* Conclusions */

2013-03-20T16:39:53Z

Conclusions

← Older revision		Revision as of 16:39, 20 March 2013
Line 141:		Line 141:

	=Conclusions=		=Conclusions=
	The compression simulation could be a useful tool for image forensics workers to determine how many times an image has been compressed. It correctly simulates the Fourier artifacts that arise after different levels of compression. There are limitations to this software, though. As is pointed out in reference 1, if the second compression is evenly divisible by the first compression, nothing is detected. Also, if the uncompressed image has periodic coefficients to begin with (a rare but possible case), a detection scheme might get a false positive.		The compression simulation could be a useful tool for image forensics workers to determine how many times an image has been compressed. It correctly simulates the Fourier artifacts that arise after different levels of compression. We can imagine a situation in which a Forensic analyst sees some random histogram of DCT coefficients and suspects some compression scheme. He could very easily apply the suspected compression scheme (or any other) to an uncompressed image to see if the Histograms have similar Fourier transforms. There are limitations to this software, though. As is pointed out in reference 1, if the second compression is evenly divisible by the first compression, nothing is detected. Also, if the uncompressed image has periodic coefficients to begin with (a rare but possible case), a detection scheme might get a false positive.

imported>Projects221: /* Image Forensics */

2013-03-20T16:37:18Z

Image Forensics

← Older revision		Revision as of 16:37, 20 March 2013
Line 3:		Line 3:

	=Image Forensics=		=Image Forensics=
	In a forensic analyst's ideal world, the images we see would be faithful representations of what was seen by the camera. There are situations in which we would like to detect whether an image has been JPEG compressed multiple times. A doubly JPEG compressed image would indicated that the image has been saved multiple times and is thus more likely to have been tampered with in between the first save (ostensibly by the camera) and the second save (perhaps in photoshop or some other editing software.) Based on the work of Dartmouth Professor Hany Farid (among others), this project ~~attempts to implement~~ code that ~~will detect doubly compressed images~~. ~~There are~~ a ~~other ways to go about detecting image forgeries~~. ~~One is~~ to implement ~~an Expectation/Maximization algorithm that can detect and interpolate~~ a ~~camera's color filter array. Parts of an~~ image ~~that do not fit in well with~~ the ~~periodic nature of the CFA will come under suspicion as forged. This project, though, will take advantage~~ of JPEG ~~compression algorithm (and the artifacts it leaves behind) to detect image forgeries~~.		In a forensic analyst's ideal world, the images we see would be faithful representations of what was seen by the camera. There are situations in which we would like to detect whether an image has been JPEG compressed multiple times. A doubly JPEG compressed image would indicated that the image has been saved multiple times and is thus more likely to have been tampered with in between the first save (ostensibly by the camera) and the second save (perhaps in photoshop or some other editing software.) Based on the work of Dartmouth Professor Hany Farid (among others), this project implements code that simulates any level of compression. Thus, when a forensic analyst looks at the DCT coefficients from a JPEG file, he can compare these coefficients with any suspected multiple compression scheme.

			I also attempted to implement a tester to look at a JPEG image but was foiled by the task of ripping DCT coefficients directly from JPEG.

	=Introduction=		=Introduction=
	Let us begin with the ideas behind JPEG compression. JPEG (which stands for Joint Photographic Experts Group) has become an image compression standard. The algorithm behind standard JPEG compression is as follows:		Let us begin with the ideas behind JPEG compression. JPEG (which stands for Joint Photographic Experts Group) has become an image compression standard. The algorithm behind standard JPEG compression is as follows:

imported>Projects221: /* Methods */

2013-03-20T16:34:53Z

Methods

← Older revision		Revision as of 16:34, 20 March 2013
Line 62:		Line 62:

	=Methods=		=Methods=
			This derivation for two compressions is somewhat accessible, but let us say we want to develop something that detects higher levels of compression, we need to know what we are looking for. This is why I developed something to simulate any number of compression levels. It allows us to almost instantly see empirical evidence for what the Fourier transforms of the compression would look like. This would be an invaluable tool in understanding what kind of compressions an image has undergone.

	==Simulating Compression==		==Simulating Compression==
	The code in Appendix A simulates multiple compressions on an uncompressed image:		The first attached code in Appendix A simulates multiple compressions on an uncompressed image:

	The general method is this:		The general method is this:

imported>Projects221: /* Results of Simulating Double Compression */

2013-03-20T15:49:27Z

Results of Simulating Double Compression

← Older revision		Revision as of 15:49, 20 March 2013
Line 123:		Line 123:

	[[File:24_double_comp.png]]		[[File:24_double_comp.png]]


			It was also found that histogram shapes and Fourier shapes stayed the same when converting the image back and forth betweeen RGB and YCbCr formats in matlab.

	==Results of Simulating Triple Compression==		==Results of Simulating Triple Compression==

imported>Projects221: /* Appendix 1 */

2013-03-20T10:20:07Z

Appendix 1

← Older revision		Revision as of 10:20, 20 March 2013
Line 148:		Line 148:

	=Appendix 1=		=Appendix 1=
			Code for simulating multiple compression
			[[File:Multi_Comp.zip]]


			Inconclusive code for detecting patterns in decompressed coefficients
			[[File:Testing.zip]]

imported>Projects221: /* Conclusions */

2013-03-20T10:19:06Z

Conclusions

← Older revision		Revision as of 10:19, 20 March 2013
Line 136:		Line 136:


	I would have liked to create a tool that takes in an image and determines if it has been compressed multiple times. I attempted to leverage the fact that the decompressed coefficients have some periodic nature, but this periodicity seemed inconclusive as far as I could tell. If one could replace the image reading and coefficient extraction with direct coefficient extraction from the .jpg file, I am confident that this would work within the bounds described above in results (i.e. b/a not equal to integer)		I would have liked to create a tool that takes in an image and determines if it has been compressed multiple times. I attempted to leverage the fact that the decompressed coefficients have some periodic nature, but this periodicity seemed inconclusive as far as I could tell. If one could replace the image reading and coefficient extraction with direct coefficient extraction from the .jpg file, I am confident that this would work within the bounds described above in results (i.e. b/a not equal to integer). This could also be done by working with a given quantization table. The roundabout method would be to decompress in matlab and recompress by the given quantization table.

	If I were to continue working on this, I would try to extract the dct coefficients directly from the .jpg file (I found this very difficult for some reason). Professor Hany Farid supplied me with some .mex code to do it, but I could not get that to compile, nor could I find any way to access the .jpg file directly. Someone who has a better low-level understanding of computers would have better luck.		If I were to continue working on this, I would try to extract the dct coefficients directly from the .jpg file (I found this very difficult for some reason). Professor Hany Farid supplied me with some .mex code to do it, but I could not get that to compile, nor could I find any way to access the .jpg file directly. Someone who has a better low-level understanding of computers would have better luck.

imported>Projects221: /* Appendix 1 */

2013-03-20T10:11:20Z

Appendix 1

← Older revision		Revision as of 10:11, 20 March 2013
Line 148:		Line 148:

	=Appendix 1=		=Appendix 1=
	~~[Simulation_Script]~~



	~~[Testing_Script]~~
	~~[Function_Style_Code]~~

imported>Projects221: /* Conclusions */

2013-03-20T09:58:54Z

Conclusions

← Older revision		Revision as of 09:58, 20 March 2013
Line 136:		Line 136:


	I would have liked to create a tool that takes in an image and determines if it has been compressed multiple times. I attempted to leverage the fact that the decompressed coefficients have some periodic nature, but this periodicity seemed inconclusive as far as I could tell.		I would have liked to create a tool that takes in an image and determines if it has been compressed multiple times. I attempted to leverage the fact that the decompressed coefficients have some periodic nature, but this periodicity seemed inconclusive as far as I could tell. If one could replace the image reading and coefficient extraction with direct coefficient extraction from the .jpg file, I am confident that this would work within the bounds described above in results (i.e. b/a not equal to integer)

	If I were to continue working on this, I would try to extract the dct coefficients directly from the .jpg file (I found this very difficult for some reason). Professor Hany Farid supplied me with some .mex code to do it, but I could not get that to compile, nor could I find any way to access the .jpg file directly. Someone who has a better low-level understanding of computers would have better luck.		If I were to continue working on this, I would try to extract the dct coefficients directly from the .jpg file (I found this very difficult for some reason). Professor Hany Farid supplied me with some .mex code to do it, but I could not get that to compile, nor could I find any way to access the .jpg file directly. Someone who has a better low-level understanding of computers would have better luck.

imported>Projects221: /* References */

2013-03-20T09:57:29Z

References

← Older revision		Revision as of 09:57, 20 March 2013
Line 141:		Line 141:

	=References=		=References=
	~~Shower Praise on Hany~~		H. Farid, Dartmouth Department of Computer Science, Lectures at Stanford, Winter, 2013

			[1] A.C. Popescu and H Farid "Statistical Tools for Digital Forensics." Provided by Stanford Psychology 221. Winter,2013

			[2]H. Farid. "Image Forgery Detection: A Survey" IEEE Signal Processing Magazine. March 2009: 16-25

	=Appendix 1=		=Appendix 1=
	[Simulation_Script]		[Simulation_Script]