Research On Information Hiding In Speech Perception Authentication System

With the rapid development of information technology such as computer science technology and communication technology, the multimedia information, represented by speech, not only enriches people’s daily life, but also brings a lot of information security issues in these years, such as illegal alteration and malicious attacks. These problems cause the rise of speech content integrity authentication research of which representative is speech perceptual hashing authentication system. In the framework of speech perceptual hashing authentication, combining with information hiding technology, this thesis mainly research on the embedding and hiding of perceptual hash value in primitive domain and mobile computing environment, including transparency, robustness and embedding capacity, and embedding rate. Specific contents of this thesis are as follows:1. In accordance with the problem that there is no scheme of perceptual hash value embedding or hiding in speech perceptual hashing authentication, a watermarking algorithm, based on bipolar quantization and phase coding, which is suitable for speech perceptual hashing authentication and can transmit perceptual hash value efficiently, is proposed in this thesis. At the transmitter, the algorithm embeds watermarks in embedding positions scrambled by Logistic mapping using improved phase coding by bipolar quantization and reconstructs the speech signal. At the receiver, the algorithm extracts watermarks based on quantitative identification after embedding positions determination. Experimental results show that the proposed scheme has not only good transparency and robustness but also security. The algorithm proposed is satisfied with the requirements of perceptual speech hashing authentication. To improve embedding capacity of watermarking algorithm used in speech perceptual hashing authentication, based on algorithm proposed above, a watermarking algorithm using modified bipolar quantization which can achieve high capacity is proposed in this thesis. This algorithm modifies bipolar quantization and doubles the capacity of the previous watermarking algorithm. Experimental results show that, compared with the previous one, the proposed algorithm improves the embedding capacity and ensures transparency and robustness. It can satisfy capacity requirements of most of perceptual hashing algorithms.2. In accordance with the problem that how to hide perceptual hash value efficiently in mobile environment, based on G.729 codec and digital-position hiding method, an information hiding scheme is proposed in this thesis. This scheme embeds perceptual hash values in the least significant bits of G.729 frames located by the perceptual evaluation of the speech quality(PESQ) which are selected to carry secret data. Experimental results show that the proposed scheme was of good transparency, higher embedding efficiency, and superb algorithm efficiency that satisfies the real-time performance requirement of the speech perceptual hashing authentication for G.729 coded speech.