Digital Forensic Technology For Audio Forgery Detection

Due to the popularization of portable recording devices,it tends to be that people can obtain digital audio materials conveniently and it's much easier to edit audio files with the powerful multimedia editing software.If these tampered audios were used as electronic evidence in court or other important occasions,it will likely lead to certain undesirable social issues.Therefore,digital audio forgery detection technology plays a significant role in verifying the credibility of digital audio data and it will be widely used in judicial forensics and news authenticity authentication.Digital audio forgery includes content forgery and property forgery,this paper primarily focuses on research of audio content forgery detection technology.Two types of detection methods to respectively detect copy-move forgery in the same audio and splicing forgery are given.The main work of this paper is as follows:1.The general detection procedures of audio copy-move forgery and audio splicing forgery are presented.Two methods for audio forgery detection were introduced and the shortcomings of the two methods are indicated by simulating experiments.2.For audio copy-move forgery,a detection method based on DTW algorithm is proposed.First,we use a pitch tracking method to exact pitch sequences of every word in the audio,after calculating DTW distance between every two words' pitch sequence and comparing the DTW distance with the threshold,we can detect the copy-move forgery in the audio.Besides,this thesis builds a speech database,in which 2 persons(1 male and 1 female),2 recording devices and 4 recording environments are involved.From the experimental results,we can find that the method has a better performance compared with the same type of algorithms under noise environment.3.For audio splicing forgery,a detection method based on silent segment is proposed.First,we extract MFCC feature which expressed the information of the recording device and expand the feature by GLDS-kernel function,after calculating correlation coefficients by a sliding window and located the position of splicing by the abnormal of the correlation coefficients.Experimental results show that the method can effectively detect the splicing forgery and resist to addictive noise,resample and compression.