Fooling the Watchers: Breaking AIGC Detectors via Semantic Prompt Attacks
This work addresses security concerns for platforms and users relying on AIGC detectors to prevent identity misuse, though it is incremental in improving adversarial attack methods.
The authors tackled the problem of AI-generated content (AIGC) detectors being vulnerable to adversarial attacks by proposing an automated framework that generates semantic prompts to evade detection, achieving top performance in a real-world competition.
The rise of text-to-image (T2I) models has enabled the synthesis of photorealistic human portraits, raising serious concerns about identity misuse and the robustness of AIGC detectors. In this work, we propose an automated adversarial prompt generation framework that leverages a grammar tree structure and a variant of the Monte Carlo tree search algorithm to systematically explore the semantic prompt space. Our method generates diverse, controllable prompts that consistently evade both open-source and commercial AIGC detectors. Extensive experiments across multiple T2I models validate its effectiveness, and the approach ranked first in a real-world adversarial AIGC detection competition. Beyond attack scenarios, our method can also be used to construct high-quality adversarial datasets, providing valuable resources for training and evaluating more robust AIGC detection and defense systems.