A New Steganalysis Method for Adaptive Spread Spectrum Steganography

In this paper we focus on the detection of spread spectrum steganography, especially in the case that the scaling parameters are adaptive to image content. Our contribution is two folds. Firstly, we present a new steganalysis method for adaptive spread spectrum steganography: block-based scatter difference detection. Our algorithm first restores cover image by restoring filter. Then the embedding process is simulated in both test and cover images for certain times and the statistic in local DCT block is calculated as a criterion for classification. Secondly, we propose a region estimation algorithm to estimate the most likely stego regions. Our experiment result is promising by showing hits over 90% and false alarm lower than 10%. Its simplicity and extensibility indicate its further application in watermark detection.