| Summary: | This PhD thesis provides new research results in the area of using 3D features for steganalysis. The research study presented in the thesis proposes new sets of 3D features, greatly extending the previously proposed features. The proposed steganlytic feature set includes features representing the vertex normal, curvature ratio, Gaussian curvature, the edge and vertex position of the 3D objects in the spherical coordinate system. Through a second contribution, this thesis presents a 3D wavelet multiresolution analysis-based steganalytic method. The proposed method extracts the 3D steganalytic features from meshes of different resolutions. The third contribution proposes a robustness and relevance-based feature selection method for solving the cover-source mismatch problem in 3D steganalysis. This method selects those 3D features that are robust to the variation of the cover source, while preserving the relevance of such features to the class label. All the proposed methods are applied for identifying stego-meshes produced by several steganographic algorithms.
|
|---|
