With the growing attention on multimedia security, various MP3 steganographic and steganalytic algorithms have been proposed increasingly. However, the existing MP3 steganalysis is lack of good universality and detection performance. To address this problem, we devise an effective MP3 steganalytic algorithm based on joint point-wise and block-wise correlations of quantified modified discrete cosine transfer coefficients matrix. On the one hand, a universal rich high pass filtering module for MP3 steganalysis is deployed to boost the sensitiveness of the algorithm to the subtle signal brought by the data hiding. On the other hand, based on the principle of MP3 encoding and the characteristics of MP3 steganography, multi-scale correlations measure module is introduced, which is used to measure the changes of point-wise, 2 × 2 block-wise, and 4 × 4 block-wise correlations due to steganography separately. Additionally, customized feature optimization, including truncated threshold selection and high pass filters measure, is performed based on the properties of MP3 steganography. Experimental results illustrate that our proposed algorithm can be applied to various MP3 steganographic algorithms, bitrates, duration, and relative payloads. Detection accuracies are promoted by more than 20% averagely compared with state-of-the-art methods.
Read full abstract