Secure Semi-fragile Watermarking Algorithm Based On Audio Content

With the rapid development of the digital information media and the computer network technology, it is easily to edit, transmit, produce and tamper digital audio files, which make copyright ownership and content authentication of digital audio to be an urgent problem to solve. This thesis mainly does research on watermark generate methods and tamper detection based on semi-fragile watermarking with audio content as research object.First, this thesis introduces the research background and significance of the digital audio watermarking briefly, next is the basic knowledge of digital audio watermarking. Then, it analysis and conclude watermarking from three aspects namely:the watermark generation, watermark embedding and tamper detection, which could be considered as the basis of the algorithm proposed in this thesis.In order to improve the security of semi-fragile audio watermarking scheme, a secure semi-fragile audio watermarking algorithm based on the approximate components energy (ACE) is proposed in this thesis. In this scheme, every audio frame is divided into two sections. One is used to extract the feature of audio frame, and the other is used to embed the watermark data of other audio frame. The feature of an audio frame is the logarithm of ACE of the chosen audio section form the audio frame to the base a. For each audio frame, the feature is encrypted and embedded in hybrid domain of another audio frame based on the secret key randomly. The validity of an audio frame is determined by the watermark inconsistency of itself and its neighborhood audio frames. This paper also discusses the value of a based on the robustness and distribution of the audio frame feature. Experimental results demonstrate that the proposed scheme can localize the tampered regions and resist collage attack effectively.This thesis proposes a secure semi-fragile audio watermarking scheme based on SVD, which could solve the problems in present semi-fragile audio watermarking algorithms as follows:small watermarking capacity, high false detecting and leaking-detecting probability, and so on. In this scheme, every audio frame is divided into two sections to generate and embed watermarking. For audio section to generate watermarking, it extracts features by singular decomposition, generates watermarking by key and gray code based on its characteristics, and then embeds watermarking on the singular value of other audio frames randomly. In tamper detecting, this scheme judges whether there is malicious tampering by bits and threshold of extracted watermark inconsistent with refactoring watermark. Then coupled with the frames may be tampered in the adjacent audio frames to determine whether tampering occurred. The experiment results show that this algorithm can accurately locate tampering and collage attack, and has lower false detecting and leaking-detecting probability...