An evaluation of wavelet filters performance for steganalysis

This paper presents an evaluation of wavelet filters performance for the task of steganalysis. We analyzed six different wavelet filters namely Daubachies, Coiflets, Symlets, Discrete Meyer, Biorthogonal and Reverse Biorthogonal families for feature extraction in a wavelet based steganalysis technique. Two publicly available steganography tools, namely the F5 steganography and the model based steganography were used to embed messages in a database of clean images to develop steganographic database of images. A Fisher linear discriminant classifier is trained using all six feature sets extracted from both clean and steganographic images separately and subsequently used for classification. Experiments revealed that the features using the dasiaHaar(db1)psila wavelet filter gave the best steganalysis performance.