Intermediate Representations are Strong AI-Generated Image Detectors

TL;DR

This paper introduces a search-based detection method using intermediate layer embeddings to identify AI-generated images, outperforming existing training-based and training-free techniques across benchmarks.

Contribution

The authors propose a novel, computationally efficient detection approach leveraging embedding sensitivity, improving generalization and performance over prior methods.

Findings

Achieves 39.61% AUROC improvement on Forensics Small over training-free methods.

Outperforms state-of-the-art detection techniques on GenImage and Forensics Small benchmarks.

Demonstrates robustness across different datasets and unseen data domains.

Abstract

The rapid advancement in generative AI models has enabled the creation of photorealistic images. At the same time, there are growing concerns about the potential misuse and dangers of generated content, as well as a pressing need for effective AI-generated image detectors. However, current training-based detection techniques are typically computationally costly and can hardly be generalized to unseen data domains, while training-free methods fall short in detection performance. To bridge this gap, we propose a search-based method employing data embedding sensitivity in intermediate layers to detect AI-generated images. Given a set of real and AI-generated images, our method examines the similarity between original image embeddings and perturbed image embeddings, and detects AI-generated images based on the similarity. We examine the proposed method on two comprehensive benchmarks: GenImage and Forensics Small. Our method exhibits improved performance across different datasets compared to both training-free and training-based state-of-the-art methods. On average, our method achieves the largest performance gain on the Forensics Small benchmark by 39.61% compared to the best training-free method and 5.14% compared to the best training-based method in AUROC score.