Tamper Detection And Tamper Recovery Based On Voice Watermarking

With the rapid development of digital technology and multimedia technology,speech signals have become the main carrier of daily information transmission.Since the speech signal is easily distorted by malicious tampering during transmission,if the tampering of the speech signal involves a series of important data such as national security,court evidence,medicine,etc.,the tampered signal will cause serious consequences to the society.Therefore,more and more attention has been paid to the security of speech signals.As an important branch of information hiding technology,speech watermarking can effectively protect speech signals by embedding information into speech signals.Because of its concealment and robustness,speech watermarking is widely used in copyright protection,tampering hints and information encryption.This paper makes an in-depth study of speech watermarking algorithm from two aspects:(1)The proposal and application of tamper detection algorithm based on speech watermarking.Aiming at the problem that traditional tamper detection algorithms cannot withstand de-synchronization attack,this paper proposes a dual authentication algorithm on the premise of ensuring the watermark concealment and robustness,which can solve the synchronization problem of the watermark and effectively detect the tampering.First,we extract the most significant bit(MSB)from each frame to generate its synchronization code as the first authentication information.Then,the local characteristic coefficient and global characteristic coefficient are calculated from each frame signal,and the two characteristics are combined together as the secondary authentication information.Through experimental simulation of various common tamper attacks,the proposed tamper detection algorithm is evaluated.Experimental results show that the proposed method not only has high tamper detection accuracy,but also ensures the watermark has good inaudibility and robustness.(2)The proposal and application of tampering and recovery algorithm based on speech watermarking.Speech watermark algorithm can not only detect tamper,but also recover the tampered content.To realize tamper recovery,a large amount of recovery information needs to be embedded into the speech signal,but it will inevitably destroy the quality of the original speech signal.Therefore,how to balance the inaudibility of watermark and the recovery ability of watermark is a chal enging problem.In this paper,the least significant bit(LSB)substitution embedding method is adopted to provide as much embedding space as possible while ensuring the quality of the original speech,and G723.1 encoding is adopted to compress and restore the information to reduce the amount of embedded watermark.In the process of watermark embedding,the authentication information and the recovery information are embedded by using the aligned strategy and the unaligned strategy respectively to ensure the watermark's recovery ability.The experimental results show that the damaged speech can be recovered according to the proposed method even if most of the original speech is tampered.In addition,the quality of recovered speech is satisfactory.