DCT Statistics and Pixel Correlation based Blind Image Steganalysis

Abstract

Steganography is an art of secret communication which is used to hide secret message into cover objects. With the growing need for information security, various steganography techniques have evolved. Due to advancements in internet technology and trends of online sharing on the social network, threats of steganography techniques for the malicious purpose have also increased. The availability of a large number of free steganography tools on the internet makes it simpler to accomplish these negative steganography activities. In order to combat the threats due to such malicious steganography, there is need of a technique known as steganalysis which can detect such malicious communications. Any form of digital media like text, image, audio, and video can be used as a carrier for secret communication. Now a days, digital images are most commonly used on social networks. This is the reason why many researchers have proposed to use image as a carrier for steganography. Hence establishing strong need of image steganalysis and it s research. This thesis discusses this assertion in detail and describes the theoretical and experimental work done to show its validity. newlineThe existing state-of-art image steganalysis methods make use of higher dimensional feature set which increases the timing complexity of the classifier. The use of various image formats and message types used in steganography have created a necessity to analyze the performance of image steganalysis across various image formats, image types and hidden message types. Existing steganalysis methods are mostly designed to break either spatial, transform or spread spectrum steganography domains but not all of them. Moreover, the performance of existing steganalysis methods is not found satisfactory when dealing with low embedding rate steganography images. Detailed literature review revealed that there is a requirement to develop lower dimensional image steganalysis algorithm which can break unknown image steganography methods from any domain with variable embedding