Layer Consistency Matters: Elegant Latent Transition Discrepancy for Generalizable Synthetic Image Detection

TL;DR

This paper introduces a novel method called latent transition discrepancy (LTD) that leverages inter-layer consistency differences in latent representations to improve the detection of synthetic images, enhancing generalization and robustness.

Contribution

The paper presents a new approach that captures inter-layer transition discrepancies to distinguish real from synthetic images, outperforming existing methods in accuracy and generalizability.

Findings

LTD exceeds baseline models by 14.35% in mean accuracy across datasets.

LTD outperforms recent state-of-the-art detection methods.

The approach demonstrates superior robustness and generalization in synthetic image detection.

Abstract

Recent rapid advancement of generative models has significantly improved the fidelity and accessibility of AI-generated synthetic images. While enabling various innovative applications, the unprecedented realism of these synthetics makes them increasingly indistinguishable from authentic photographs, posing serious security risks, such as media credibility and content manipulation. Although extensive efforts have been dedicated to detecting synthetic images, most existing approaches suffer from poor generalization to unseen data due to their reliance on model-specific artifacts or low-level statistical cues. In this work, we identify a previously unexplored distinction that real images maintain consistent semantic attention and structural coherence in their latent representations, exhibiting more stable feature transitions across network layers, whereas synthetic ones present discernible distinct patterns. Therefore, we propose a novel approach termed latent transition discrepancy (LTD), which captures the inter-layer consistency differences of real and synthetic images. LTD adaptively identifies the most discriminative layers and assesses the transition discrepancies across layers. Benefiting from the proposed inter-layer discriminative modeling, our approach exceeds the base model by 14.35\% in mean Acc across three datasets containing diverse GANs and DMs. Extensive experiments demonstrate that LTD outperforms recent state-of-the-art methods, achieving superior detection accuracy, generalizability, and robustness. The code is available at https://github.com/yywencs/LTD