Leveraging Failed Samples: A Few-Shot and Training-Free Framework for Generalized Deepfake Detection

TL;DR

This paper introduces FTNet, a training-free few-shot deepfake detection method that leverages a small number of samples for improved real-world detection, achieving state-of-the-art results without model training.

Contribution

The work presents a novel training-free few-shot framework for deepfake detection that uses only one fake sample during evaluation, differing from traditional training-based approaches.

Findings

Achieves new state-of-the-art performance on 29 generative models.

Improves detection accuracy by an average of 8.7% over existing methods.

Demonstrates effectiveness of leveraging failed samples in few-shot scenarios.

Abstract

Recent deepfake detection studies often treat unseen sample detection as a ``zero-shot" task, training on images generated by known models but generalizing to unknown ones. A key real-world challenge arises when a model performs poorly on unknown samples, yet these samples remain available for analysis. This highlights that it should be approached as a ``few-shot" task, where effectively utilizing a small number of samples can lead to significant improvement. Unlike typical few-shot tasks focused on semantic understanding, deepfake detection prioritizes image realism, which closely mirrors real-world distributions. In this work, we propose the Few-shot Training-free Network (FTNet) for real-world few-shot deepfake detection. Simple yet effective, FTNet differs from traditional methods that rely on large-scale known data for training. Instead, FTNet uses only one fake samplefrom an evaluation set, mimicking the scenario where new samples emerge in the real world and can be gathered for use, without any training or parameter updates. During evaluation, each test sample is compared to the known fake and real samples, and it is classified based on the category of the nearest sample. We conduct a comprehensive analysis of AI-generated images from 29 different generative models and achieve a new SoTA performance, with an average improvement of 8.7\% compared to existing methods. This work introduces a fresh perspective on real-world deepfake detection: when the model struggles to generalize on a few-shot sample, leveraging the failed samples leads to better performance.