LAA-X: Unified Localized Artifact Attention for Quality-Agnostic and Generalizable Face Forgery Detection

2026-04-05

AI Summary

This work addresses the limited generalization of existing deepfake detection methods under high-quality and unseen manipulation scenarios. To this end, the authors propose the LAA-X framework, which leverages multi-task learning and a hybrid data synthesis strategy to explicitly guide the model toward local regions prone to generating forgery artifacts. Central to this approach is a local artifact attention mechanism, complemented by an auxiliary task designed to enhance sensitivity to vulnerable facial regions. The framework is compatible with both CNN- and Transformer-based backbones, instantiated as LAA-Net and LAA-Former, respectively. Trained solely on real and forged samples without additional supervision, LAA-X achieves state-of-the-art or competitive performance across multiple benchmarks, demonstrating significantly improved robustness in detecting high-fidelity and previously unseen fake faces.

Abstract

In this paper, we propose Localized Artifact Attention X (LAA-X), a novel deepfake detection framework that is both robust to high-quality forgeries and capable of generalizing to unseen manipulations. Existing approaches typically rely on binary classifiers coupled with implicit attention mechanisms, which often fail to generalize beyond known manipulations. In contrast, LAA-X introduces an explicit attention strategy based on a multi-task learning framework combined with blending-based data synthesis. Auxiliary tasks are designed to guide the model toward localized, artifact-prone (i.e., vulnerable) regions. The proposed framework is compatible with both CNN and transformer backbones, resulting in two different versions, namely, LAA-Net and LAA-Former, respectively. Despite being trained only on real and pseudo-fake samples, LAA-X competes with state-of-the-art methods across multiple benchmarks. Code and pre-trained weights for LAA-Net\footnote{https://github.com/10Ring/LAA-Net} and LAA-Former\footnote{https://github.com/10Ring/LAA-Former} are publicly available.

Problem

• face forgery detection
• generalization
• high-quality forgeries
• unseen manipulations
• deepfake detection

Innovation

• Localized Artifact Attention
• Multi-task Learning
• Blending-based Data Synthesis
• Quality-Agnostic Detection
• Generalizable Face Forgery Detection