Audio content forgery detection in broadcasting is crucial to prevent misinformation spreading. Tools for the authentication of audio files can be proven very useful, and several techniques have been proposed. In the current paper a database for evaluation of such techniques is introduced. A script was created for automatic generation of tampered audio files, given a number of original source files that contain recorded speech, while they have been encoded in different audio formats (Mp3, AAC, AMR, FLAC) and bitrates and finally they were used to generate the tampered audio files. The database was subjectively evaluated by experts in terms of samples changing audibility. The effect of tampering on several audio features was tested, in order to propose semi-automatic methods for discrimination between the original and tampered files. The database and the scripts are publically accessible so that researchers can use the pre-generated files or use the script to create datasets oriented to their research interests.
(See document for exact affiliation information.)
https://aes2.org/publications/elibrary-page/?id=19403
(1205KB)
Click to purchase paper as a non-member or login as an AES member. If your company or school subscribes to the E-Library then switch to the institutional version. If you are not an AES member Join the AES. If you need to check your member status, login to the Member Portal.
