FairAdapter: Detecting AI-generated Images with Improved Fairness

TL;DR

FairAdapter is a new framework designed to improve the fairness of AI-generated image detection, ensuring consistent performance across diverse image contents, surpassing existing methods.

Contribution

We introduce FairAdapter, a novel approach that enhances detection fairness for AI-generated images, addressing overfitting issues in current deep learning forensic models.

Findings

Achieves higher fairness in detection across varied image contents

Outperforms state-of-the-art methods in fairness metrics

Demonstrates robust detection performance on diverse datasets

Abstract

The high-quality, realistic images generated by generative models pose significant challenges for exposing them.So far, data-driven deep neural networks have been justified as the most efficient forensics tools for the challenges. However, they may be over-fitted to certain semantics, resulting in considerable inconsistency in detection performance across different contents of generated samples. It could be regarded as an issue of detection fairness. In this paper, we propose a novel framework named Fairadapter to tackle the issue. In comparison with existing state-of-the-art methods, our model achieves improved fairness performance. Our project: https://github.com/AppleDogDog/FairnessDetection