Steganalysis for MP3Stego using differential statistics of quantization step

MP3Stego is a typical steganographic tool for MP3 audio, which embeds secret message into MP3 audio according to the parity of the block length. In this paper, a detection method for MP3Stego based on the differential statistics of quantization step is present. By analyzing the algorithm of MP3Stego, we find that the quantization step which is an important parameter of MP3 encoding is affected during embedding. The standard deviation of the second-order differential sequence from quantization step is adopted as the classification feature. Experimental results indicate that the proposed algorithm is effective in detecting MP3Stego and can achieve better detection performance than the existing algorithms.

► The effect on the encoder, which is caused by the MP3Stego embedding is analyzed.
► The features extracted show good discriminatory ability even at low embedding rates.
► Differential process can increase the featureʼs sensitivity.