Abstract: MP3 is the most popular audio format nowadays in our daily life, for example music downloaded from the Internet and file saved in the digital recorder are often in MP3 format. However, low bitrate MP3s are often transcoded to high bitrate since high bitrate ones are of high commercial value. Also audio recording in digital recorder can be doctored easily by pervasive audio editing software. In this paper, we investigate the distributions of the first digits of MDCT coefficients of audio, and propose two methods to detect double compression of audio. The methods are essential for finding out fake-quality MP3 and audio forensics. The proposed methods use support vector machine classifiers with feature vectors formed by the distributions of the first digits of the quantized MDCT (modified discrete cosine transform) coefficients. Experimental results demonstrate that both methods can discriminate double-compressed audio from single-compressed audio with error rate approximating 0 when BR2 is larger than BR1. Even when BR2 is less than BR1, the method of band division can detect double compression with error rate below 20\%. To the best of our knowledge, this piece of work is the first one to detect double compression of audio signal.

@inproceedings{yang-shi-huang-spie10,
  url          = {http://link.aip.org/link/?PSISDG/7541/75410K/1},
  booktitle    = {SPIE Conference on Media Forensics and Security},
  author       = {Rui Yang and Yun Qing Shi and Jiwu Huang},
  location     = {San Jose, CA},
  year         = {2010},
  title        = {Detecting double compression of audio signal},
}