Steganalysis based on Markov model and SVM for JPEG images
Cong Huang · Journal of Computer Applications · 2007
A new and universal steganalysis scheme based on Markov model for JPEG images was proposed. 360 dimensional feature vectors were derived from the markov matrix, which was calculated directly in DCT domain and sensitive to the data embedding process. Then Support Vector Machine (SVM) was used to classify. The experimental results of 4 popular JPEG steganographic schemes (F5, Outguess, MB1 and MB2) demonstrate that the proposed scheme has the advantage in detection rate, stability, and speed.