Speech Content Authentication Based On Digital Watermarking

With the recent development of the Network technology and multimedia technology,various multimedia products have brought a lot of conveniences to human life.Similarly,it also leads to some risks and challenges that seem to be obstacles to overcome.As a special kind of multimedia data,speech signal always contains private content and sensitive information,such as instructions in military commanding system,card number and password,and the identity information of speaker.It is no doubt that the data owners do not want their sensitive speech to be eavesdropped by untrusted third party.If those sensitive speech signals are subjected to malicious tampering,it will probably bring about immeasurable loss(such as personal property,reputation),or even worse,it could as well jeopardize national security.Therefore,privacy-protection and integrity authentication of digital speech signal are fundamental issues to be resolved.In this thesis,the studies are focused on the content integrity authentication of digital speech signal in plaintext domain and encrypted domain.The primary contributions are listed as follows.1.In order to locate the tampered area of speech files accurately and to judge the tampered type,a dual fragile watermarking algorithm is proposed based on crisscross method.In the proposed scheme,a host speech signal is firstly transformed into a matrix,then elements of each row and each column of that matrix are combined with a sensitive hash function to generate two fragile watermarks(row fragile watermark and column fragile watermark).In the process of tampering detection,if one or more sampling points of that watermarked speech are modified,the reconstructed row and column watermarks will be changed.Therefore,the tampered sampling points can be located by those error crisscross points.Experimental results show that the proposed algorithm can not only detect attacks,but also accurately locate the tampered samples,and even identify the type of counterfeits.2.A G.723.1-based speech content authentication algorithm by modifying the weight of least significant digits is proposed.Firstly,the G.723.1 coding technology is applied to generate the compression speech signal which is combined with frame number and check code to form a watermark.Additionally,a novel watermark embedding scheme by modifying the weight of least significant digits of original speech is introduced to guarantee the inaudibility of watermark.Experimental results show that the proposed watermarking algorithm is extremely fragile.Meanwhile,it has favorable inaudibility and accurate tampering locationcapability.Additionally,those tampered speech frames can be restored with high-quality to improve the utilization rate of that attacked speech.3.We propose a content authentication scheme in encrypted speech with restoration capabilities by combining encryption and watermarking.In that algorithm,hyper-chaotic system is introduced to generate multi-dimension pseudo-random sequences for hiding the feature and improving the confidentiality of original speech.Also,a compression matrix is used to produce the compression information.When a receiver acquires an attacked speech signal,the tampered area can be located and recovered by the extracted check-bits and self-embedded auxiliary compression bits,respectively.Analysis and experimental results demonstrate that the proposed algorithm provide a high security,and the self-embedding watermark can locate and recover the content of that tampered speech.4.An efficient speech watermarking algorithm is proposed for content authentication and recovery in encrypted domain.In the proposed algorithm,chaotic and block cipher are combined to eliminate the positional correlation and conceal the statistical feature for protecting the privacy and confidentiality.Moreover,those tampered sampling points of encrypted speech can be recovered by using the correlation between their adjacent samples.Simulation results show that the proposed scheme has high security to protect the privacy of speech.Furthermore,it also has strong ability to locate and recover the tampered area without using any auxiliary information.