Forensic reserve method improves image forgery detection with minimal training
Forensic Reserve: Eliciting Latent Knowledge for Image Forgery Detection
Computer Vision and Pattern Recognition
Summary
As fake images become harder to spot, it’s important to have better ways to detect them. The authors show that pretrained vision models already contain hidden clues useful for spotting forgeries. They developed a method called Reserve-Guided Elicitation (RGE) that finds and strengthens these clues without needing lots of new training data. This helps create lightweight detectors that work well on different tests without extra tuning.
What this means in practice
- •For security software developers: Create lightweight forgery detectors using pretrained models with minimal additional training data for reliable fake image detection.
- •For digital content verification teams: Improve detection accuracy of forged images by incorporating forensic reserve adaptations without retraining full vision models.
Authors
Jiahua Li, Zixu John, Tom Zhong, Fuping Wu, Tianhao Xu, Jianqing Zheng, Yuanhan Mo, Fei Shen
Abstract
As generated images become increasingly realistic, reliable forgery detection is essential for maintaining trust in visual information. However, existing methods primarily rely on task-specific supervision to adapt vision foundation model representations, without fully exploiting internal forensic knowledge to guide detection. To address this limitation, we propose Reserve-Guided Elicitation (RGE), a framework that treats sparse, origin-sensitive internal components in pretrained models as a forensic reserve and translates their localization into structural constraints for lightweight adaptation. Specifically, we first use the Forensic Lens (F-lens) to decompose activations across layers and token groups into independent components and globally screen them by their response differences between real and generated images, identifying reserve sites and directions. Next, we map the selected directions back to hidden-state space to construct fixed reserve subspaces and insert Forensic Reserve Adapters (FRA) only at the identified sites. Finally, with the backbone parameters, previously fitted reference classifier, and subspace bases fixed, we train only the FRA coefficient maps to generate input-dependent residual updates constrained to the corresponding subspaces, strengthening existing forensic responses. Using only 500 labeled training images and a trainable parameter budget below 0.2% of the backbone, RGE achieves competitive performance across three detection benchmarks without target-benchmark adaptation. Furthermore, RGE consistently improves over the corresponding frozen detectors across eight encoders spanning self-supervised and vision-language pretraining, eliciting a latent forensic capacity broadly shared across pretrained vision models.