I-Corps: Development of an Audio Forensic Analysis Tool

Regents Of The University Of Michigan - Ann Arbor, Ann Arbor MI

Investigators

Abstract

The use of digital multimedia (images, audio, & video) as evidence is rapidly growing in multiple civic applications, including legal proceedings and law enforcement. Verifying the authenticity and integrity of digital evidence is a complex and a challenging task given the ease with which digital information can be copied and altered using widely available editing tools. In particular, in the absence of helping data (i.e., digital watermarks) and if the only input is the media itself, authentication poses unique problems for digital forensics. Until now existing state-of-the-art is unable to uniquely link a given digital recording to "the" microphone, link an audio recording to the location it was made, or detect traces of anti-forensic processing. The proposed audio forensic analysis tool enables a forensics analyst to provide authoritative answers to basic forensic investigation questions. The project aims to demonstrate effectiveness of the audio forensic analysis tool capable to detecting forgeries in digital audio recordings. The proposed tool relies on acoustic cues, microphone nonlinearity traces, and post-processing artifacts left in the digital audio recording. The proposed tool uses statistical data modeling and feature extraction techniques to capture these fingerprints, and use them for linking a given recording to "the" acquisition device and to "the" acoustic environment. This team believes that the proposed audio forensic analysis tool, when successfully commercialized, could provide forensic examiners, traditional media, news outlets, social media, and law enforcement personnel a very powerful, computationally efficient, scalable, and reliable tool to perform online (as well as offline) forensics analysis, to determine authenticity of the digital media object, localize forgery (if any) in the digital media object.