Beyond Semantics: Uncovering the Physics of Fakes via Universal Physical Descriptors for Cross-Modal Synthetic Detection

TL;DR

This paper identifies stable physical features that distinguish AI-generated images from real ones across diverse datasets and integrates them into multimodal models like CLIP to improve synthetic image detection.

Contribution

It introduces a novel feature selection algorithm for robust physical features and demonstrates their integration into CLIP for state-of-the-art detection performance.

Findings

Identified five core physical features with consistent discriminative power.

Achieved 99.8% accuracy on multiple Genimage benchmarks.

Enhanced detection by combining pixel-level features with semantic information.

Abstract

The rapid advancement of AI generated content (AIGC) has blurred the boundaries between real and synthetic images, exposing the limitations of existing deepfake detectors that often overfit to specific generative models. This adaptability crisis calls for a fundamental reexamination of the intrinsic physical characteristics that distinguish natural from AI-generated images. In this paper, we address two critical research questions: (1) What physical features can stably and robustly discriminate AI generated images across diverse datasets and generative architectures? (2) Can these objective pixel-level features be integrated into multimodal models like CLIP to enhance detection performance while mitigating the unreliability of language-based information? To answer these questions, we conduct a comprehensive exploration of 15 physical features across more than 20 datasets generated by various GANs and diffusion models. We propose a novel feature selection algorithm that identifies five core physical features including Laplacian variance, Sobel statistics, and residual noise variance that exhibit consistent discriminative power across all tested datasets. These features are then converted into text encoded values and integrated with semantic captions to guide image text representation learning in CLIP. Extensive experiments demonstrate that our method achieves state-of-the-art performance on multiple Genimage benchmarks, with near-perfect accuracy (99.8%) on datasets such as Wukong and SDv1.4. By bridging pixel level authenticity with semantic understanding, this work pioneers the use of physically grounded features for trustworthy vision language modeling and opens new directions for mitigating hallucinations and textual inaccuracies in large multimodal models.