Breaking Permutation-Based Mesh Steganography and Security Improvement

Permutation-based steganography in polygonal meshes can provide considerably large embedding capacities for hiding secret messages. However, corresponding steganalysis techniques against such methods have never been studied. This paper identifies the essential differences between naturally generated meshes and meshes produced by permutation-based steganography methods. It is found that the two types of meshes differ significantly in the distribution of topological distances between consecutive mesh elements. Therefore, by measuring the orderliness of the vertex list and the face list of meshes, we develop solutions for several mesh steganalysis problems. These solutions are effective, leading to high detection accuracy; and they are also universal, requiring no knowledge such as which steganography method is used and what data embedding rate is adopted for the detection mechanism to work. Moreover, this paper also presents a security-improved permutation-based mesh steganography approach, by taking advantage of the connectivity information of polygonal meshes and establishing a good trade-off between embedding capacity and undetectability. Without bringing global changes, our approach embeds secret messages into local neighborhoods on meshes. As a result, meshes generated by the proposed steganography approach tend to have natural structures that are unlikely to draw suspicions to steganalyzers.