MozJPEG is a JPEG compression library that aims to reduce file size while preserving perceived image quality. The library improves the rate-distortion tradeoff with trellis optimization, which modifies the Discrete Cosine Transform (DCT) coefficients of an image. Steganography is the art and science of hiding secret messages in inconspicuous covers, such as JPEG images. The goal of steganalysis is to detect the presence of steganographically hidden information. So far, the steganalysis community has devoted little attention to the implications of using different JPEG compression libraries. The uptake of MozJPEG calls for a revisit of established methods that have been developed and tested before trellis optimization was commonly used.
The objective of the thesis is to re-evaluate selected detectors of steganography in the JPEG domain (e.g., Jsteg, F5, J-UNIWARD). The student will implement the steganalysis methods and measure how sensitive they are to the use of trellis optimization. Ideally, the method can be adapted to reduce in particular the number of false alarms in scenarios where covers compressed with MozJPEG are present.