Cell-Phone Identification from Recompressed Audio Recordings

Abstract

Many audio forensic applications would benefit from the ability to classify audio recordings, based on characteristics of the originating device, particularly in social media platforms where an enormous amount of data is posted every day. This paper utilizes passive signatures associated with the recording devices, as extracted from recorded audio itself, in the absence of any extrinsic security mechanism such as digital watermarking, to identify the source cell-phone of recorded audio. It uses device-specific information present in low as well as high-frequency regions of the recorded audio. On the only publicly available dataset in this field, MOBIPHONE, the proposed system gives a closed set accuracy of 97.2 % which matches the state of art accuracy reported for this dataset. On audio recordings which have undergone double compression, as typically happens for a recording posted on social media, the proposed system outperforms the existing methods (4% improvement in average accuracy).