BENet: A Cross-domain Robust Network for Detecting Face Forgeries via Bias Expansion and Latent-space Attention

View on arXiv ← Back to list

Authors: Weihua Liu, Jianhua Qiu, Said Boumaraf, Chaochao lin, Pan liyuan, Lin Li, Mohammed Bennamoun, Naoufel Werghi

Published: 2024-12-10 11:41:55+00:00

AI Summary

This paper introduces BENet, a Cross-Domain Robust Bias Expansion Network designed for enhanced detection of fake faces across diverse deepfake generation techniques. BENet utilizes a bias expansion module, based on autoencoders, to amplify differences in fake reconstructions while preserving genuine facial features. Additionally, it incorporates a Latent-Space Attention (LSA) module for capturing multi-scale inconsistencies and a cross-domain detector to improve recognition accuracy for unknown deepfake sources.

Abstract

In response to the growing threat of deepfake technology, we introduce BENet, a Cross-Domain Robust Bias Expansion Network. BENet enhances the detection of fake faces by addressing limitations in current detectors related to variations across different types of fake face generation techniques, where ``cross-domain refers to the diverse range of these deepfakes, each considered a separate domain. BENet's core feature is a bias expansion module based on autoencoders. This module maintains genuine facial features while enhancing differences in fake reconstructions, creating a reliable bias for detecting fake faces across various deepfake domains. We also introduce a Latent-Space Attention (LSA) module to capture inconsistencies related to fake faces at different scales, ensuring robust defense against advanced deepfake techniques. The enriched LSA feature maps are multiplied with the expanded bias to create a versatile feature space optimized for subtle forgeries detection. To improve its ability to detect fake faces from unknown sources, BENet integrates a cross-domain detector module that enhances recognition accuracy by verifying the facial domain during inference. We train our network end-to-end with a novel bias expansion loss, adopted for the first time, in face forgery detection. Extensive experiments covering both intra and cross-dataset demonstrate BENet's superiority over current state-of-the-art solutions.

Key findings

BENet demonstrates superior performance over state-of-the-art solutions in both intra-dataset and cross-dataset evaluations across various benchmarks, including FF++, Celeb-DF, DFFD, and DFDC. It achieves high accuracy and AUC scores, significantly outperforming competitors in challenging scenarios. The network also exhibits strong robustness against unseen perturbations and generalizes well to novel manipulation techniques.

Approach

BENet solves face forgery detection by employing a bias expansion module (an autoencoder) that differentiates real and fake faces by expanding the bias in their reconstructions. It further integrates a Latent-Space Attention (LSA) module to capture multi-scale, forgery-related inconsistencies in the latent space. Finally, a cross-domain detector module is activated during inference to verify facial domains and enhance detection accuracy against unknown deepfake attacks.

Datasets

FaceForensics++ (FF++), Celeb-DF, Diverse Fake Face Dataset (DFFD), DeepFake Detection Challenge Dataset (DFDC)

Model(s)

UNKNOWN

Author countries

UNKNOWN

← Previous