Supervised Contrastive Learning for Few-Shot AI-Generated Image Detection and Attribution

2025-11-20

AI Summary

The rapid advancement and increasing photorealism of generative AI image models render conventional detection methods—relying on periodic retraining—computationally prohibitive and impractical for deployment. Method: This paper proposes a two-stage open-set detection framework: (1) a vision encoder is trained via supervised contrastive learning to extract discriminative embeddings; (2) a k-nearest neighbors (k-NN) classifier enables few-shot cross-model generalization for both detection and provenance attribution, requiring no fine-tuning to adapt to unseen generators. Contribution/Results: Under a stringent few-shot setting with only 150 images per class, the method achieves 91.3% detection accuracy (+5.2% over baselines), while improving provenance AUC and open-set classification rate (OSCR) by 14.70% and 4.27%, respectively. It significantly enhances robustness against unknown generative models and enables reliable source inference without model-specific adaptation.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Generation

Application Category

• Social Networks and Social Media: Generative AI / large language models and their impact on social systems
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Security and Privacy: Data transparency and provenance

Abstract

The rapid advancement of generative artificial intelligence has enabled the creation of synthetic images that are increasingly indistinguishable from authentic content, posing significant challenges for digital media integrity. This problem is compounded by the accelerated release cycle of novel generative models, which renders traditional detection approaches (reliant on periodic retraining) computationally infeasible and operationally impractical. This work proposes a novel two-stage detection framework designed to address the generalization challenge inherent in synthetic image detection. The first stage employs a vision deep learning model trained via supervised contrastive learning to extract discriminative embeddings from input imagery. Critically, this model was trained on a strategically partitioned subset of available generators, with specific architectures withheld from training to rigorously ablate cross-generator generalization capabilities. The second stage utilizes a k-nearest neighbors (k-NN) classifier operating on the learned embedding space, trained in a few-shot learning paradigm incorporating limited samples from previously unseen test generators. With merely 150 images per class in the few-shot learning regime, which are easily obtainable from current generation models, the proposed framework achieves an average detection accuracy of 91.3%, representing a 5.2 percentage point improvement over existing approaches . For the source attribution task, the proposed approach obtains improvements of of 14.70% and 4.27% in AUC and OSCR respectively on an open set classification context, marking a significant advancement toward robust, scalable forensic attribution systems capable of adapting to the evolving generative AI landscape without requiring exhaustive retraining protocols.

Problem

• Detecting AI-generated images that are indistinguishable from authentic content
• Addressing generalization challenges with rapidly evolving generative models
• Developing scalable forensic systems without exhaustive retraining requirements

Innovation

• Two-stage detection framework with supervised contrastive learning
• k-NN classifier using few-shot learning paradigm
• Strategic generator partitioning to test generalization capabilities