关键词: 加性噪声模型, 音频隐写分析, 小波包分解, 自适应预测, 支持向量机

Abstract:

An audio steganalysis method based on wavelet packet and adaptive predictor is proposed. In this scheme, audio signals are firstly decomposed by the wavelet packet, the wavelet packet coefficients of high frequency are then predicted by the LMS adaptive predictor, statistics of predicted errors are selected as the statistical features, and finally SVM is implemented as a classifier. The experimental results verify that, for the commonly used steganography methods of additive noise, high classification accuracy can be achieved even in the case of low embedding strength or low embedding rate.

Key words: additive noise model, audio steganalysis, wavelet packet decomposition, adaptive predictor, support vector machine