Hierarchical Representation Network for Steganalysis of QIM Steganography in Low-Bit-Rate Speech Signals

Abstract

With the volume of Voice over IP (VoIP) traffic rising shapely, more and more VoIP-based steganography methods have emerged in recent years, which poses a great threat to the security of cyberspace. Low bit-rate speech codecs are widely used in the VoIP application due to its powerful compression capability. Previous steganalysis methods mostly focus on capturing the inter-frame correlation or intra-frame correlation features in code-words ignoring the hierarchical structure which exists in speech frame. In this paper, motivated by the complex multi-scale structure, we design a Hierarchical Representation Network (HRN) to tackle the steganalysis of Quantization Index Modulation (QIM) steganography in low-bit-rate speech signal. In the proposed model, Convolution Neural Network (CNN) is used to model the hierarchical structure in the speech frame, and three levels of attention mechanisms are applied at different convolution blocks, enabling it to attend differentially to more and less important contents in speech frame. Experiments demonstrated that the steganalysis performance of the proposed method outperform the state-of-the-art methods especially in detecting both short and low embeded speech samples. Moreover, our model needs less computation and has higher time efficiency to be applied to real online services.