Forgery Detection Of Digital Audio

With the rapid development and low-cost of digital multimedia, digital multimedia applications become more and more widely. Digital multimedia has many advantages: easy to edit, easy to copy and easy to transmit. But these advantages bring also many problems.Digital audio and video materials store in binary form. By low-cost and easy methods, it's very easy to perform forgeries to digital audio and video. This has destroyed information security of digital audio and video materials, especially in some special scene. For example, Judicial evidence, confidential documents, medical records, historical documents backup, etc. It is undoubtedly very significant to ensure the authenticity and integrity of such information.Digital audio and video materials in court as electronic evidence, as an important justice application, we must ensure the authenticity. In this paper, we focus on digital audio and address the problem of authenticating digital audio signals assuming no explicit prior knowledge of the original in conjunction with forgery point where digital audio materials are generated and transmitted.Base on clarity of authenticate elements of electronic evidence, deep understanding of format sand contents of digital audio, digital audio forgery tools and methods, from the audio analysis of Verifier-Tuple theory, we research on three main aspects: audio waveform statistical characteristics, incidental background noise characteristics, audio additional information.Audio waveform statistical characteristics: base on high order spectral analysis methods, detect forgeries using bispectral analysis. Use matlab for experiments and perform analysis on the result.Incidental background noise characteristics: From the sort and character of background noise, analyze the generation and composition of background noise. Use wavelet theory to denoise and to get the device-associated background noise as one characteristic to detect digital audio forgeries. And research on noise consistency base on blind estimation.Audio additional information: base on audio encoding and decoding, detect digital forgeries by abnormal additional information.May this paper one good reference for future work of digital audio forgery detection.