Self-Supervised Learning for Detecting AI-Generated Faces as Anomalies

TL;DR

This paper introduces a self-supervised anomaly detection approach for identifying AI-generated faces by learning face-specific features from photographic images and modeling their distribution, enabling detection of novel AI face generators.

Contribution

The paper proposes a novel self-supervised method that leverages camera-intrinsic and face features to detect AI-generated faces without retraining for new generators.

Findings

Effective detection of AI-generated faces demonstrated through experiments

Outperforms traditional binary classifiers in generalizing to new AI face generators

Utilizes a pretext task based on ranking EXIF tags and classifying manipulated images

Abstract

The detection of AI-generated faces is commonly approached as a binary classification task. Nevertheless, the resulting detectors frequently struggle to adapt to novel AI face generators, which evolve rapidly. In this paper, we describe an anomaly detection method for AI-generated faces by leveraging self-supervised learning of camera-intrinsic and face-specific features purely from photographic face images. The success of our method lies in designing a pretext task that trains a feature extractor to rank four ordinal exchangeable image file format (EXIF) tags and classify artificially manipulated face images. Subsequently, we model the learned feature distribution of photographic face images using a Gaussian mixture model. Faces with low likelihoods are flagged as AI-generated. Both quantitative and qualitative experiments validate the effectiveness of our method. Our code is available at \url{https://github.com/MZMMSEC/AIGFD_EXIF.git}.