Study On The Audio Forgery Detection Under Three Cases

As the rapid development of multimedia technology, the digital information has been penetrated every corner of our life.In the field of audio technology, it's more convenient to acquire the audio information than ever before, and people can edit or modify the audio easily through the audio editing software.Those conscious or unconscious tampering actions will pose a great threat to the security of the audio information. As the great demand for applying the digital information to forensics, digital audio forgery detection technique becomes an internationally hot topic present.In this paper, the main task focuses on digital audio forgery detection.Three innovative detection methods are proposed respectively according to the double compression tampering, the coping tampering and the audio re-sample tampering.(1)Aiming at the double compression tampering manner, a spectrum analysis detection method is proposed. In the AMR encoding and decoding process, the audio signals must be filtered by the high-pass filter with 80Hz cut-off frequency, due to the ideal filter can't be realized, the low frequency components of the second AMR audio are weakened seriously. Utilizing this characteristic, spectrum analysis detection method is proposed. This method can detect the forgery audio of AMR format by comparing the frequency component in low level.(2) Aiming at the coping and pasting tampering manner, a method based on fingerprinting compare is proposed. Utilizing the great difference between the different words, this method can detect the coping temper in the same audio segment by extracting the better robust fingerprinting. Furthermore, according to the coping or cutting process by the Cool Edit, a detecting method based on the characteristic of Cool Edit is proposed. This method has a high ability to distinguish the authenticity of the audio and detect the editing manner (cut or paste).(3) A method based on comparing frequency components in high level is proposed for the audio re-sample tampering. Based on the analysis linear interpolation re-sampling, a conclusion is obtained:the sequence formed by original points has higher zero-crossing rate than the sequence formed by interpolation points.So, the linear interpolation re-sample tampering can be detected through comparing frequency components in high level.Experimental results show that all of the proposed methods can detect efficiently according to their own tampering manners.