Improved Forensic and Anti-Forensic Techniques for JPEG Compressed Images

Abstract

Digital image forensics aims to evaluate the image authenticity by analyzing its processing history. This history relates to the origin, credibility and steps of processing that it has experienced and helps the detective to find the truth of an image. JPEG is a regularly utilized compression standard and it has been broadly utilized in cameras and image processing software’s. Therefore, JPEG compression has become an important part of many image forgeries. Thus, the detection of JPEG compression can add a great value to evaluate the authenticity of digital images. Therefore, there is need of efficient forensic techniques to evaluate the authenticity and proficient antiforensic techniques that challenge and help in the upgradation of forensic techniques are required. The research work is directed to design an efficient JPEG anti-forensic technique that has the capability to mislead the forensic detectors by hiding the artifacts of compression in Discrete Cosine Transform (DCT) domain. In the first stage of the proposed scheme, shifted block DCT approach is employed on the considered JPEG compressed image to fill the gaps in the comb-like distribution of DCT coefficients. This shifted block DCT approach led to the addition of dithering noise itself without the need of any adaptive dithering model. The result of this shifted block DCT approach is further processed in the second stage by TV (Total variation)-based deblocking operation to remove the blocking artifacts left during the JPEG compression in the spatial domain. The experimental results illustrate that the presented approach has better performance in comparison to the existing techniques in terms of image visual quality and forensic undetectability with highly reduced computational cost. 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. Therefore, the technique which has the capability to deceive the scalar based and machine learning-based forensic detectors by hiding the artifacts of compression in DFrCT domain is proposed. The additional fractional parameter ′𝛼′ in DFrCT gives more flexibility when designing a system, in contrast to DCT. The shifted block DFrCT approach efficiently hides the JPEG compression artifacts and disguises several JPEG forensic detectors. The proposed technique possesses the advantage of having an additional fractional parameter to increase its flexibility. TV-based deblocking is further used to reduce the blocking artifacts. Therefore, this technique possesses different variables for each considered image i.e. shifted block and fractional parameter for the optimization of proposed v JPEG anti-forensic approach. It is observed from the experimental results that the proposed approach provides improved performance in terms of image quality and forensic undetectability when compared to existing techniques. The goal 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 antiforensics conceals the artifacts of JPEG compression. Moreover, the analysis of JPEG antiforensics reveals the limitations of existing forensic detectors. Actually, the existing forensic work on JPEG compression detection is solely dedicated to the first-order statistical feature components analysis. The first order statistical analysis based forensic detectors can be easily misguided by applying some anti-forensic techniques. Therefore, higher order statistical analysis is required to counter these anti-forensic techniques. 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 Markov Transition Probability Matrices (MTPMs) in DCT and DFrCT domain. The proposed framework comprises of three stages: Selection of the target difference image, Evaluation of MTPMs, and Generation of second-order statistical feature based on MTPMs. In the first stage, we explore the effects of dithering operation of JPEG anti-forensics by analyzing the variance inconsistencies along the diagonals. Afterwards, MTPMs 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 MTPMs is very efficient even in the presence of anti-forensic attacks. Moreover, the multi-purpose nature of the proposed counter JPEG anti-forensic scheme is confirmed when evaluated on spliced images considering CASIA v1.0 and Columbia datasets provides better results in the detection of these image operations. The ability of the proposed forensic technique is authenticated from the extensive experimental analysis which provides better detection results against various anti-forensic techniques in terms of minimum decision error, when compared to the existing techniques. The further research work can be concentrated to design a multi-purpose forensic technique based on deep learning by considering Convolutional Neural Networks (CNN) in order to detect the different image processing operations.