Order within Chaos: Capturing Intrinsic Energy Anomalies for AI-Manipulated Image Forgery Localization

2026-06-01

AI Summary

This work addresses the challenge of detecting and localizing photorealistic images synthesized by generative AI, which often lack conventional physical noise traces exploited by existing forensic methods. The authors propose FLAME, a novel framework that reveals an intrinsic energy anomaly in diffusion-generated images—specifically, suppressed local high-frequency variance—and captures it via a Local Anomaly Detection (LAD) map. To enable precise pixel-level localization, they design a lightweight adapter to enhance the Segment Anything Model (SAM). Recognizing the rapid evolution of generative models, they further introduce EditStream, an instruction-driven data synthesis pipeline for scalable training data generation. Evaluated across multiple AI-generated image benchmarks, FLAME achieves state-of-the-art performance and demonstrates strong generalization to unseen generative architectures.

Abstract

Recent advancements in generative AI have led to image editing models capable of producing realistic forgeries that evade traditional image forgery localization methods, as these approaches depend on physical noise absent in synthetic data. To address this challenge, we theoretically demonstrate that the diffusion process inherently suppresses local high-frequency variance, creating a statistical energy gap that is distinguishable from the natural entropy of optical imaging. Guided by this insight, we propose FLAME, a unified framework that utilizes a LAD map to capture these intrinsic anomalies, coupled with a parameter-efficient adapter for SAM to achieve precise, pixel-level forgery localization. Furthermore, to bridge the lag between forensic benchmarks and evolving generative models, we introduce EditStream, an automated pipeline for continuous, instruction-based training data synthesis. Extensive experiments demonstrate that FLAME establishes a new state-of-the-art, significantly outperforming previous methods on AI-generated forgery datasets while effectively generalizing to unseen generative architectures. Our code is available at https://github.com/phoenixnir/FLAME.

Problem

• image forgery localization
• generative AI
• synthetic image detection
• forensic analysis
• AI-generated images

Innovation

• diffusion process
• energy anomaly
• forgery localization
• parameter-efficient adapter
• automated data synthesis