Improving Interpretability and Robustness for the Detection of AI-Generated Images

TL;DR

This paper enhances the robustness of AI-generated image detection by analyzing existing methods, proposing embedding and attention head improvements, and introducing a new dataset for better generalization across models.

Contribution

It introduces novel methods to improve detection robustness and provides a new dataset to facilitate future research in AI-generated image detection.

Findings

Increased out-of-distribution classification scores by up to 6%.

Analysis of CLIP-based detection methods reveals interpretability insights.

New dataset supports cross-model generalization evaluation.

Abstract

With growing abilities of generative models, artificial content detection becomes an increasingly important and difficult task. However, all popular approaches to this problem suffer from poor generalization across domains and generative models. In this work, we focus on the robustness of AI-generated image (AIGI) detectors. We analyze existing state-of-the-art AIGI detection methods based on frozen CLIP embeddings and show how to interpret them, shedding light on how images produced by various AI generators differ from real ones. Next we propose two ways to improve robustness: based on removing harmful components of the embedding vector and based on selecting the best performing attention heads in the image encoder model. Our methods increase the mean out-of-distribution (OOD) classification score by up to 6% for cross-model transfer. We also propose a new dataset for AIGI detection and use it in our evaluation; we believe this dataset will help boost further research. The dataset and code are provided as a supplement.