Circumventing shortcuts in audio-visual deepfake detection datasets with unsupervised learning

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Smeu, Stefan, Boldisor, Dragos-Alexandru, Oneata, Dan, Oneata, Elisabeta
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913865639395328
author Smeu, Stefan
Boldisor, Dragos-Alexandru
Oneata, Dan
Oneata, Elisabeta
author_facet Smeu, Stefan
Boldisor, Dragos-Alexandru
Oneata, Dan
Oneata, Elisabeta
contents Good datasets are essential for developing and benchmarking any machine learning system. Their importance is even more extreme for safety critical applications such as deepfake detection - the focus of this paper. Here we reveal that two of the most widely used audio-video deepfake datasets suffer from a previously unidentified spurious feature: the leading silence. Fake videos start with a very brief moment of silence and based on this feature alone, we can separate the real and fake samples almost perfectly. As such, previous audio-only and audio-video models exploit the presence of silence in the fake videos and consequently perform worse when the leading silence is removed. To circumvent latching on such unwanted artifact and possibly other unrevealed ones we propose a shift from supervised to unsupervised learning by training models exclusively on real data. We show that by aligning self-supervised audio-video representations we remove the risk of relying on dataset-specific biases and improve robustness in deepfake detection.
format Preprint
id arxiv_https___arxiv_org_abs_2412_00175
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Circumventing shortcuts in audio-visual deepfake detection datasets with unsupervised learning
Smeu, Stefan
Boldisor, Dragos-Alexandru
Oneata, Dan
Oneata, Elisabeta
Computer Vision and Pattern Recognition
Machine Learning
Sound
Audio and Speech Processing
Image and Video Processing
Good datasets are essential for developing and benchmarking any machine learning system. Their importance is even more extreme for safety critical applications such as deepfake detection - the focus of this paper. Here we reveal that two of the most widely used audio-video deepfake datasets suffer from a previously unidentified spurious feature: the leading silence. Fake videos start with a very brief moment of silence and based on this feature alone, we can separate the real and fake samples almost perfectly. As such, previous audio-only and audio-video models exploit the presence of silence in the fake videos and consequently perform worse when the leading silence is removed. To circumvent latching on such unwanted artifact and possibly other unrevealed ones we propose a shift from supervised to unsupervised learning by training models exclusively on real data. We show that by aligning self-supervised audio-video representations we remove the risk of relying on dataset-specific biases and improve robustness in deepfake detection.
title Circumventing shortcuts in audio-visual deepfake detection datasets with unsupervised learning
topic Computer Vision and Pattern Recognition
Machine Learning
Sound
Audio and Speech Processing
Image and Video Processing
url https://arxiv.org/abs/2412.00175