An Improved Secret Key Estimation Method for Sequential Steganography

This paper presents an improved algorithm to detect the beginning and end of the secret message during low frequency embedding in sequential steganography. Based on the fact that the statistics of stego signal have a higher jump after sequential embedding, the method effectively indicates the jump in low frequency embedding through adjusting the log likelihood estimator, namely, increasing the power of the standard deviations and introducing the modulating factors, and then develops a new cumulative sum detector. Experiments show that this method still can estimate the secret key in the low frequency embedding even though the message strength is as low as 0.3.