Statistical Approaches for Digital Image Steganalysis

Abstract

Steganography is the technique to hide secret information within cover objects like images, audio, video and text files. It has been widely reported that there has been a surge in the use of steganography for criminal activities and therefore, implementing effective detection techniques is an essential task in digital forensics. Unfortunately, building a single effective detection technique still remains one of the biggest challenges. The proliferation of steganographic tools has created a demand for powerful means to detect hidden data. This thesis presents three steganalysis techniques which are developed using statistical properties of an image. When secret data is hidden in an image, the statistical properties like variance, correlation, entropy, PSNR, and MSE are changed due to the hidden secret data. We have used these quantitative measures to detect whether any secret data is present in the image or not. Using a statistical approach, we investigated the inherent detectability of several commonly used steganography techniques to check the performance of proposed steganalysis approaches.