Copy-Move Forgery Detection and Localization via Partial Audio Matching

In this paper, we present a new approach for detecting and localizing copy-move forgeries within digital audio recordings. The approach is based on an audio fingerprinting and matching algorithm that was originally designed for partial audio matching without a query for large datasets, which was now successfully adapted to the specific requirements of copy-move forgery detection, i.e. short segment duration and high reliability. Thanks to the characteristics of the original algorithm, our proposed approach does not require pre-segmentation, and shows high accuracy for detection and localization of copy move forgery, including mismatching background noise, thereby significantly extending the state-of-the art.