Detecting Audio Deepfakes on the Edge:Lightweight SSL-Based Detection in a Browser Plugin

2026-06-29

AI Summary

This study addresses the privacy risks associated with cloud-based audio deepfake detection, which hinder secure and localized verification needed by journalists and fact-checkers. To overcome this limitation, the authors propose a lightweight on-device detection method that, for the first time, integrates a truncated self-supervised learning (SSL) backbone with a logistic regression classifier, substantially reducing model complexity. The approach maintains high privacy while achieving a 10% improvement in accuracy over the AASIST baseline and a 40% increase in inference speed. Furthermore, the method has been successfully deployed into a publicly accessible browser extension, enabling rapid and private verification of audio authenticity directly on users’ devices.

Abstract

Audio deepfakes are a growing challenge for the general public, as well as for journalists and fact-checkers. The latter need reliable tools to verify the authenticity of their sources, while at the same time keeping their information private. Commercial deepfake detection solutions rely on cloud-based processing, which raises privacy concerns. To solve this problem, we propose an on-device audio deepfake detection model. We show that a truncated self-supervised backbone with a simple logistic classifier is both very fast and often more accurate than existing solutions. Our solution outperforms the baseline AASIST by 10% and improves inference speed by 40%. We integrate this model into a browser plug-in, which allows journalists and fact-checkers to detect deepfakes easily and securely. Code for the plugin is available at https://github.com/OctavianPascu97/Audio-Deepfakes-Browser-Plugin.

Problem

• Audio Deepfakes
• Privacy
• Edge Detection
• Browser Plugin
• On-device Processing

Innovation

• on-device detection
• self-supervised learning
• audio deepfake
• browser plugin
• lightweight model