CAMME: Adaptive Deepfake Image Detection with Multi-Modal Cross-Attention

TL;DR

CAMME introduces a multi-modal cross-attention framework that significantly improves deepfake detection accuracy and robustness across unseen generative models and adversarial attacks by integrating visual, textual, and frequency features.

Contribution

The paper presents CAMME, a novel multi-modal cross-attention approach that enhances deepfake detection generalization and robustness against unseen architectures and adversarial perturbations.

Findings

Achieves over 12% improvement in detection accuracy on natural scenes.

Maintains over 91% accuracy under natural image perturbations.

Reaches 89.01% and 96.14% accuracy against PGD and FGSM attacks.

Abstract

The proliferation of sophisticated AI-generated deepfakes poses critical challenges for digital media authentication and societal security. While existing detection methods perform well within specific generative domains, they exhibit significant performance degradation when applied to manipulations produced by unseen architectures--a fundamental limitation as generative technologies rapidly evolve. We propose CAMME (Cross-Attention Multi-Modal Embeddings), a framework that dynamically integrates visual, textual, and frequency-domain features through a multi-head cross-attention mechanism to establish robust cross-domain generalization. Extensive experiments demonstrate CAMME's superiority over state-of-the-art methods, yielding improvements of 12.56% on natural scenes and 13.25% on facial deepfakes. The framework demonstrates exceptional resilience, maintaining (over 91%) accuracy under natural image perturbations and achieving 89.01% and 96.14% accuracy against PGD and FGSM adversarial attacks, respectively. Our findings validate that integrating complementary modalities through cross-attention enables more effective decision boundary realignment for reliable deepfake detection across heterogeneous generative architectures.