Please use this identifier to cite or link to this item:
Full metadata record
DC FieldValueLanguage
dc.contributor.supervisorSingh, Kulbir-
dc.contributor.authorSingh, Gurinder-
dc.description.abstractThe digital image information can be easily manipulated without leaving any footprints due to the availability of powerful image processing tools. Thus, there is an immediate need to confirm the legitimacy of digital images. Most of the cameras compress the image by employing Joint Photographic Experts Group (JPEG) standard. During the forgery creation, when this image is decompressed and re-compressed with different quantization matrix, it becomes double compressed doctored image. This JPEG double compression becomes an integral part of forgery creation. The detection and analysis of the historic information (for example, quantization matrix) related to JPEG compression in an image help the detective to find the truth of an image. The research work is directed to design a two-stage forensic technique to evaluate the first quantization matrix from the partial double compressed JPEG images. In the first stage of the proposed scheme, automatic isolation of the doubly compressed part from doctored image is performed by exploring the JPEG ghost technique. The second stage analyzes this doubly compressed part to estimate the first quantization matrix or steps. In the latter stage, an optimized filtering scheme is also proposed to cope with the effects of the error. The experiment results confirm that the first stage of the proposed scheme provides an average percentage accuracy of 95.45%. The second stage provides an error less than 1.5% for the first ten Discrete Cosine Transform (DCT) coefficients, hence, outperforming the existing techniques. The experimental results consider the partial double compressed images in which the recompression is done with different quantization matrix. The digital image forensics most often employs JPEG compression based forensic detectors. To confirm the capability of JPEG forensic detectors, an anti-forensic approach is desired. Thus, the further research is dedicated to design an enhanced JPEG anti-forensic technique in order to eliminate the blocking artifacts added during the JPEG compression in both spatial and DCT domains. In the presented approach, the grainy noise introduced in DCT domain by perceptual histogram smoothing can be reduced considerably with the application of suggested de-noising techniques. Two kinds of denoising operations are suggested, one is based on the minimization problem of Total Variation (TV) of energy and other on normalized weighted function. Afterwards, an advanced TV-based deblocking method is proposed to remove the blocking artifacts in spatial domain. Subsequently, a decalibration algorithm is employed to get back the statistics of processed image to its normal situation. The experiment results indicate that the suggested anti-forensic schemes are better than the existing methods in attaining improved tradeoff between visual quality of an image and forensic undetectability, but with high computational cost. The objective of counter JPEG anti-forensics is to expose the artifacts of JPEG compression in the presence of an anti-forensic attack. It is a challenging task because the application of JPEG anti-forensics conceals the artifacts of JPEG compression. Moreover, the analysis of JPEG anti-forensics reveals the limitations of existing forensic detectors. For example, most of the JPEG compression forensic techniques usually depend on the examination of first-order statistics based on the histogram of an image. These forensic techniques are easily circumvented by adopting an anti-forensic attack. Therefore, higher-order statistical analysis is required which is much robust against anti-forensic attacks. To resolve this issue, a counter JPEG anti-forensic approach is presented in this work by considering the second-order statistical analysis based on the Co-occurrence matrices (CMs). The proposed framework comprises of three stages: Selection of the target difference image, Evaluation of CMs, and Generation of second-order statistical feature based on CMs. In the first stage, we explore the effects of dithering operation of JPEG anti-forensics by analyzing the variance inconsistencies along the diagonals. Afterwards, CMs are evaluated in the second stage to highlight the effects of grainy noise introduced during the dithering operation. The third stage is devoted to generate an optimal second order statistical feature which is fed to the SVM classifier. The experimental results based on the UCID and BOSSBase dataset images demonstrated that the proposed forensic detector based on CM is very efficient even in the presence of anti-forensic attacks. Moreover, the proposed scheme is also evaluated in countering Median filtering and Contrast Enhancement (CE) anti-forensics. The multi-purpose nature of the proposed counter JPEG anti-forensic scheme is confirmed from the fact that it also provides better results in the detection of these anti-forensic techniques and other image operations such as Mean filtering (MeanF), Gaussian filtering (GF), Wiener filtering (WF), Scaling (Sca), and Rotation (Rot). The further research work can be concentrated to design a forensic technique based on other machine learning approaches such as Convolutional Neural Network (CNN) and Recurrent Neural Network (RNN) in order to detect the different image processing operations.en_US
dc.description.sponsorshipVisvesvaraya PHD scheme for Electronics and IT, Ministry of Electronics and Information Technology, Government of India under Grant PhDMLA/4(33)/2015-16/01en_US
dc.subjectImage tamperingen_US
dc.subjectCounter JPEG anti-forensicsen_US
dc.subjectDigital image forensicsen_US
dc.subjectDouble JPEG compressionen_US
dc.subjectJPEG anti-forensicsen_US
dc.titleDouble compressed doctored image anti-forensics with statistical forensics analysisen_US
Appears in Collections:Doctoral Theses@ECED

Files in This Item:
File Description SizeFormat 
Thesis Gurinder Singh 901406015.pdf4.42 MBAdobe PDFView/Open

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.