Statistical detection of LSB matching in the presence of nuisance parameters

This paper investigates the reliable detection of information embedded with the least significant bits (LSB) matching scheme. It is aimed to design a test with analytically predictable error probabilities. To this end, the problem of hidden information detection is cast in the framework of hypothesis testing theory. In order to deal with nuisance parameters a rejection approach is used with a statistical model of medium. The use of a linear parametric model permits to analytically express statistical performance of proposed test. Numerical simulations and comparisons with state-of-the-art detectors highlight the relevance of proposed approach.