Malicious audio source detection using audio watermarking

Audio or speech can be easily recorded and manipulated with the kind of handheld devices available today. It is important to detect and prove the authenticity of an audio source especially when it is recorded by a malicious user using a mobile phone, while the audio being played by authorized device. Detection of malicious audio source is possible by using an audio watermarking technique. This audio watermarking method must be robust against geometric distortions such as time scale modification(TSM), pitch scaling, random cropping, which are inherent when audio is played back and recorded. In this work we propose a method to find the authenticity of the source of the audio using Log Coordinate Mapping(LCM) based audio watermarking[1], which is robust to geometric distortions. The watermark bits are added in appropriate frequency bands for enabling detection of change in signal characteristics when passed through the analog channel. The empirical observations reveal that our method could robustly detect the malicious source of audio if the audio is being recorded and played back by unauthorized source.