Detection of impostor and tampered segments in audio by using an intelligent system

Mubeen, Zeshan, Afzal, Mehtab, Ali, Zulfiqar, Khan, Suleman and Imran, Muhammad (2021) Detection of impostor and tampered segments in audio by using an intelligent system. Computers & Electrical Engineering, 91. p. 107122. ISSN 0045-7906

[img] Text
COMPELECENG-D-20-01925_R2.pdf - Accepted Version
Restricted to Repository staff only until 21 March 2022.
Available under License Creative Commons Attribution Non-commercial No Derivatives 4.0.

Download (3MB) | Request a copy
Official URL: https://doi.org/10.1016/j.compeleceng.2021.107122

Abstract

The transmission of audio data via the Internet of Things makes such data vulnerable to tampering. Moreover, the availability of sophisticated tampering tools has allowed mobsters to change the context of audio data by altering their segments. Tampered audio may result in unpleasant situations for any member of society. To avoid such circumstances, a new audio forgery detection system is proposed in this study. This system can be deployed in edge devices to identify impostors and tampering in audio data. The proposed system is implemented using state-of-the-art mel-frequency cepstral coefficient features. Meanwhile, a Gaussian mixture model is used to train and validate the system. To evaluate the proposed system, a dataset of tampered audios is created by mixing recordings from two different speakers. The performance of the proposed system in authenticating genuine audio is between 92.50% and 100%, and that in detecting forged audio is between 99.90 and 100%.

Item Type: Article
Uncontrolled Keywords: Audio forgery, Splicing, Audio forensic, Zedge computing, Machine learning, Voice activity detection, Audio authentication
Subjects: G400 Computer Science
G600 Software Engineering
Department: Faculties > Engineering and Environment > Computer and Information Sciences
Depositing User: John Coen
Date Deposited: 24 Mar 2021 10:27
Last Modified: 24 Mar 2021 10:30
URI: http://nrl.northumbria.ac.uk/id/eprint/45774

Actions (login required)

View Item View Item

Downloads

Downloads per month over past year

View more statistics