Abstract:

With the development of multimedia technology and recording equipments, recording material as objective evidence increasingly appears in court trail and investigation. If the recording material becomes a judgment, it must ensure that the evidence has not been spliced or forged. In audio forensics, tamper localization is an important part of signal interpolation detection. MP3 is the most widely used digital compressed audio format currently. In Amazon.com, recording devices of top 20 sales mostly support MP3 format and a portion only support MP3 format. The default recording format of most smart mobile phones is MP3. Thus, the authenticity and the integrity of MP3 receive more widespread concern. In this paper, to solve the problem of high false positive rate of the existing method using frame offset, an improved passive algorithm in locating MP3 forgeries is proposed. Observed that the misjudgment of the existing detection algorithms often appear in samples containing more silent segments. In order to eliminate the effect of silent segments on extracting offsets, this paper uses silence detection technology to set a threshold value. Meanwhile, in order to avoid instability in extracting one frame offset, this paper uses moving window to extract. The experiment result shows, the improved method can locate deletion, insertion, substitution and splicing forgeries accurately and reduce the false positive rate largely compared with the existing method.

Key words: MP3, passive forgery detection, tamper localization, MDCT coefficients