A study on dependencies between DCT coefficients for JPEG image steganalysis

摘要

In the paper, a new blind image steganalysis scheme for JPEG image steganography is proposed. Space filling curve is exploited to scan quantized block DCT coefficients. Markov model is applied to model the dependencies between quantized DCT cofficients so as to extract intrablock and interblock transition probability matrices. The experiments on F5, Steghide and MB1 steganography show the method can detect these steganographic methods reliably and the detection accuracy exceeds Pevny's[5] obviously.