$μ$Flow: Leveraging Average Images for Improving Generalisation of Deepfake Faces Detectors

2026-06-29

AI Summary

This work addresses the limited generalization of existing deepfake detection methods across diverse generative models—such as GANs and diffusion models—and under unknown forgery scenarios. To overcome this challenge, the authors propose μFlow, a one-class unsupervised detection framework trained exclusively on authentic images. The key innovation lies in leveraging the tendency of averaged images to amplify generative artifacts; by modeling the distribution of mean features from real images and aligning individual image features to this distribution via normalizing flows, μFlow enables likelihood-based authenticity assessment. Remarkably, without requiring any forged samples or synthetic artifacts during training, μFlow achieves state-of-the-art performance in fully out-of-distribution settings, demonstrating exceptional cross-generator and cross-dataset generalization capabilities.

Abstract

Current generative models, including GANs and diffusion models, have reached an outstanding level of photorealism, posing significant risks to privacy and security. To ensure real-world applicability, deepfake detectors must generalise effectively to unseen generators. However, most existing approaches rely on supervised training with both real and fake images, which limits their generalisation especially across generators categories (e.g. GANs vs DMs). In this work, we introduce $μ$Flow, a one-class deepfake detector trained only on real images without relying on pseudo-deepfakes or synthetic artifacts. Our approach builds on the observation that averaging multiple images amplifies consistent generative traces, producing highly discriminative feature representations. We leverage this property by modelling the distribution of features extracted from averaged images and training a normalizing flow to align the feature space of individual images with this distribution. This alignment yields a likelihood-based criterion that separates real and fake samples while promoting strong generalisation. We evaluate $μ$Flow on a fully out-of-distribution setting, where both real and fake datasets are unseen during training. Experimental results show that our method significantly outperforms SOTA detectors. Project page: https://opontorno.github.io/MuFlow.

Problem

• deepfake detection
• generalisation
• out-of-distribution
• generative models
• one-class classification

Innovation

• one-class detection
• image averaging
• normalizing flow
• deepfake generalization
• generative trace amplification