Active Steganalysis of Transform Domain Steganography Based on Sparse Component Analysis
Subject Areas : Image ProcessingHamed Modaghegh 1 * , Seyed Alireza Seyedin 2
1 - Ferdowsi University of Mashhad
2 - Ferdowsi University of Mashhad
Keywords: Sparse Component Analysis (SCA), Active Steganalysis, Blind Source Separation (BSS), Transform Domain steganography, , , , ,
Abstract :
This paper presents a new active steganalysis method to break the transform domain steganography. Most of steganalysis techniques focus on detecting the presence or absence of a secret message in a cover (passive steganalysis), but in some cases we need to extract or estimate hidden message (active steganalysis). Although estimating the message is important but there is little research in this area. A new active steganalysis method which is based on Spars Component Analysis (SCA) technique is presented in this work. Here, the sparsity property of cover image and hidden message has been used to extract hidden message from stego image. In our method, transform domain steganography is formulated mathematically as a linear combination of sparse sources and therefore active steganalysis can be presented as a SCA problem. The feasibility of the SCA problem solving is confirmed by Linear Programming methods. Then, a fast algorithm is introduced to decrease the computational cost of steganalysis without much loss of accuracy. The accuracy of our new method has been confirmed in different experiments on a variety of transform domain steganography. These experiments show that, our method compared to the previous active steganalysis methods not only reduces the error rate but also decreases the computational cost.