Micro-Defects Expose Macro-Fakes: Detecting AI-Generated Images via Local Distributional Shifts

TL;DR

The paper introduces MDMF, a local distribution-aware framework that detects AI-generated images by amplifying micro-defects into macro-level discrepancies using patch-based analysis and MMD, outperforming existing methods.

Contribution

It proposes a novel local distributional shift detection method with a learnable patch forensic signature and theoretical analysis showing improved detection reliability.

Findings

MDMF outperforms baseline detectors across multiple benchmarks.

Patch-wise modeling yields larger discrepancies when forensic signals are present.

The framework effectively amplifies micro-defects into detectable macro-level differences.

Abstract

Recent generative models can produce images that appear highly realistic, raising challenges in distinguishing real and AI-generated images. Yet existing detectors based on pre-trained feature extractors tend to over-rely on global semantics, limiting sensitivity to the critical micro-defects. In this work, we propose Micro-Defects expose Macro-Fakes (MDMF), a local distribution-aware detection framework that amplifies micro-scale statistical irregularities into macro-level distributional discrepancies. To avoid localized forensic cues being diluted by plain aggregation, we introduce a learnable Patch Forensic Signature that projects semantic patch embeddings into a compact forensic latent space. We then use Maximum Mean Discrepancy (MMD) to quantify distributional discrepancies between generated and real images. Our theory-grounded analysis shows that patch-wise modeling yields provably larger discrepancies when localized forensic signals are present in generated images, enabling more reliable separation from real images. Extensive experiments demonstrate that MDMF consistently outperforms baseline detectors across multiple benchmarks, validating its general effectiveness. Project page: https://zbox1005.github.io/MDMF-project/