Deepfake Detection without Deepfakes: Generalization via Synthetic Frequency Patterns Injection

TL;DR

This paper presents a novel training method for deepfake detectors that enhances their ability to generalize across unseen generation techniques by injecting synthetic frequency patterns into pristine images.

Contribution

It introduces a technique that trains detectors with synthetic frequency patterns, improving their robustness and generalization to unknown deepfake generation methods.

Findings

Achieved state-of-the-art detection performance across 25 generation methods.

Models trained with synthetic patterns outperform previous methods in generalization.

Enhanced detection accuracy on unseen deepfake techniques.

Abstract

Deepfake detectors are typically trained on large sets of pristine and generated images, resulting in limited generalization capacity; they excel at identifying deepfakes created through methods encountered during training but struggle with those generated by unknown techniques. This paper introduces a learning approach aimed at significantly enhancing the generalization capabilities of deepfake detectors. Our method takes inspiration from the unique "fingerprints" that image generation processes consistently introduce into the frequency domain. These fingerprints manifest as structured and distinctly recognizable frequency patterns. We propose to train detectors using only pristine images injecting in part of them crafted frequency patterns, simulating the effects of various deepfake generation techniques without being specific to any. These synthetic patterns are based on generic shapes, grids, or auras. We evaluated our approach using diverse architectures across 25 different generation methods. The models trained with our approach were able to perform state-of-the-art deepfake detection, demonstrating also superior generalization capabilities in comparison with previous methods. Indeed, they are untied to any specific generation technique and can effectively identify deepfakes regardless of how they were made.