Detection of Js']Jsteg hiding using image statistical model

Jsteg is an open steganography software on Internet. It uses the LSB of DCT coefficients to hide secret information in image. Chi-square attack can detect sequential Jsteg hiding but it can't detect the random Jsteg hiding. This paper presents a new, fast steganalytic method for detecting Jsteg hiding which is more convenient than the Chi-square attack. This method can detect both the sequential Jsteg hiding and the random Jsteg hiding by using the statistical model of image. The AC coefficients of image are divided into two parts and the distribution of the statistic of the two parts is fitted by Laplacian. The mean of Laplacian is 0 that is proved by Hypothesis testing. The Pearson chi(2) test is done to test goodness-of fit. Basing on this statistic mode, our method can detect Jsteg hiding with high accuracy. The amount of embedding message can be estimated using linear regression. We test our method on hundreds of images with different amount of embedding.