Conditional Probability Based Steganalysis for JPEG Steganography

Inspired by works on the Markov process based steganalysis, we propose a new steganalysis technique based on the conditional probability statistics. Specifically we focus on its performance against the F5 software. In our experiment, we prove that the proposed technique works as well or better than the Markov process based technique in terms of classification accuracy on F5. Our main advantage is a much better computational efficiency. With different number of messages embedded, it can also be seen that the performance of steganalysis depends on the message size embedded. This paper includes the introduction to conditional probability features, how the experiment works, and the discussion of the results.