Evolving AI image detection adapts to new generators without old data
EvoKnow: Continual Knowledge Evolution for AI-Generated Image Detection
Computer Vision and Pattern RecognitionArtificial Intelligence
Summary
Detecting AI-generated images is hard because new image generators keep appearing, and detectors trained on old ones often fail on new types. The authors propose EvoKnow, a method that learns to identify new AI image styles without needing to store old images or forget past knowledge. It works by keeping a shared base understanding and adding small expert components tailored to each new generator, helping the system stay accurate as new AI images arrive. Tests show EvoKnow stays reliable over time, even with very few new examples.
What this means in practice
- •For digital forensics teams: Update AI image detection tools to recognize new generators with minimal new data, without retraining on all past data.
- •For social media platform engineers: Deploy continuous detection systems that evolve as new AI images appear, reducing manual updates and improving content moderation.
Authors
Zhiheng Peng, Wenwei Jin, Yangshi Ge, Siyu Xia, Jiawei Li, Xu Tang
Abstract
AI-generated image detectors are commonly trained on fixed generator domains and become difficult to maintain as new generative models emerge. Continual adaptation is challenging because replaying historical generated images is costly, whereas updating shared parameters with limited current-domain data can overwrite prior forensic knowledge. We propose EvoKnow, a replay-free framework that formulates continual AI-generated image detection as forensic knowledge evolution. EvoKnow preserves a shared forensic basis learned from base domains, incrementally adds isolated residual experts for complementary generator-relevant evidence, and retrieves expertise through an Analytical Incremental Router (AIR) updated in closed form from current-stage generated images and accumulated sufficient statistics. Experiments demonstrate effective cross-generator generalization, few-shot expansion, and long-horizon continual adaptation. With ten generated images per arriving generator, EvoKnow achieves 96.70% average accuracy on non-base GenImage generators and 94.48% accuracy on Chameleon without target-benchmark adaptation. Under a strict replay-free continual learning protocol, EvoKnow achieves state-of-the-art continual learning performance, attaining 96.32% mean stage-wise accuracy and 4.32% average forgetting.