Speech steganalysis using chaotic-type features

Abstract

In this paper, we investigate chaotic-type features for universal speech steganalysis. These features are used in the design of linear regression classifier. The steganalyzer is tested on watermarking and steganographic methods. Experimental results show the potential of chaotic-type features for their discriminatory power to be used in steganalysis. We believe that the integration of chaotic-type features with linear ones may capture both linear and non-linear aspects of speech signals leading to robust and efficient hybrid steganalyzers.