Abstract:

Image sequential steganography is defined as a class of embedding algorithms that hide messages in the consecutive features （time, spatial or frequency domain） of the host signal. This paper presents a steganalysis method that estimates the secret key used in the sequential steganography of spread s  spectrum embedding. We use the method of cumulative sum based on the sequential probability test （CUMUM-SPRT） to detect the changes, which is used in sequential analysis and abrupt changes detection. Considering the Laplacian distribution of image DCT coefficients, we give a secret key estimation model of stationary Laplacian host signals embedded by the spread spectrum steganography. Experiments are done using the stationary Laplacian random sequence generated by a stochastic differential equation（SDE） method. For non-statioanry digital images with a low signal noise ratio （SNR）, a locally mos st powerful steganaysis detector is derived based on the Laplacian distribution. The results of experiments show that our method is more efficient than  Trivedi＇s method.

Key words: steganalysis;secret key estimation, Laplacian distribution, image information hiding