MP3 audio steganalysis

摘要: MP3, one of the most widely used digital audio formats, provides a high compression ratio with faithful quality. The widespread use enables MP3 files to become excellent covers carry hidden information in steganography on Internet. Our research, however, indicates that there are few steganalysis methods proposed detect steganograms and for information-hiding behavior particularly scarce. In this paper we propose comprehensive approach by deriving combination features from quantized MDCT coefficients. We design frequency-based subband moment statistical features, accumulative Markov transition neighboring joint density second-order derivatives. also model distortion extracting shape parameters generalized Gaussian individual frames. Different feature selection algorithms applied improve detection accuracy. Signal complexity modification introduced provide evaluation. Experimental results show our is successful discriminating generated steganographic tool, MP3Stego, each category signal complexity, especially streams complexities generally more difficult steganalyze.