Boosting Robust AIGI Detection with LoRA-based Pairwise Training

The proliferation of highly realistic AI-Generated Image (AIGI) has necessitated the development of practical detection methods. While current AIGI detectors perform admirably on clean datasets, their detection performance frequently decreases when deployed "in the wild", where images are subjected to unpredictable, complex distortions. To resolve the critical vulnerability, we propose a novel LoRA-based Pairwise Training (LPT) strategy designed specifically to achieve robust detection for AIGI under severe distortions. The core of our strategy involves the targeted finetuning of a visual foundation model, the deliberate simulation of data distribution during the training phase, and a unique pairwise training process. Specifically, we introduce distortion and size simulations to better fit the distribution from the validation and test sets. Based on the strong visual representation capability of the visual foundation model, we finetune the model to achieve AIGI detection. The pairwise training is utilized to improve the detection via decoupling the generalization and robustness optimization. Experiments show that our approach secured the 3th placement in the NTIRE Robust AI-Generated Image Detection in the Wild challenge

Key Contributions

LoRA-based pairwise training strategy that decouples generalization and robustness optimization for AIGI detection
Distortion and size simulation techniques to match real-world test distribution
Visual foundation model fine-tuning achieving robust detection under severe distortions

🛡️ Threat Analysis

Input Manipulation Attack

The paper explicitly addresses robustness to distortions (unpredictable, complex transformations) that degrade detector performance — these distortions can be viewed as evasion attacks on detection systems. The pairwise training decouples generalization and robustness optimization, directly targeting adversarial robustness.

Output Integrity Attack

Detects AI-generated images to verify content authenticity and provenance — this is output integrity. The paper addresses the challenge of detecting synthetic content 'in the wild' under complex distortions.