Texture-Based Multiresolution Steganalytic Features for Spatial Image Steganography

Abstract

Steganography is a technique of hiding information in digital media whereas steganalysis is the detection and recovery of such hidden information. This paper focuses on extracting more efficient steganalytic features to improve the classification accuracy in spatial image steganography. Textural features are extracted using the well-known Threshold Local Binary Pattern (TLBP) method which is an extended version of a local binary pattern. A set of high-order derivative filters are incorporated before performing the TLBP operation to strengthen the stego features. Various textural features are extracted using the Grey Level Co-occurrence Matrix (GLCM), Discrete Wavelet Transform (DWT) and Contourlet Transform (CT). The ada-boost classifier is used for classification. Performance of the classifier is analyzed with various combinations of feature sets and the best performance is obtained for the combination of statistical and multi-resolution feature set. An accuracy of 93.87% is obtained.