A Known-Key Scenario for Steganalysis and a Highly Accurate Detector within It

We introduce a known-key scenario for steganalysis and propose an approach to construct highly accurate quantitative detectors within it. As an example we construct a quantitative detector based on the well-known SPAM-features and apply it to steganalysis of non-distorted bitmap images with LSB-matching. While the state-of-the-art methods aiming at solving this problem allow to detect payloads up to 0.1 bits-per-pixel, the proposed method detects effectively up to 0.001 bpp. Despite essential utilization of SPAM-features, which have been originally designed for LSB-matching, our method detects LSB-replacement as well, in average, with the same accuracy as LSB-matching. Thus, the known-key scenario is a very powerful tool, and it would be worthwhile to adapt the state-of-the-art methods to this scenario for the purpose of attaining even more accuracy of steganalyzers.