Face-trace: Open-Set Attribution and Progressive Discovery of Synthetic Face Generators

2026-07-08

AI Summary

This work addresses the challenge of source attribution for synthetic face images in open-set scenarios by proposing the first unified framework that jointly performs known generator identification, out-of-distribution sample rejection, and discovery of unknown generators through clustering. The method leverages frozen I-JEPA embeddings for classification, integrates Forensic Self-Descriptors with projected features to enable energy-based rejection, and progressively uncovers novel generators via feature fusion and clustering. Crucially, rejected samples are systematically organized rather than discarded, facilitating dynamic expansion of the generator space. Evaluated on the WILD dataset, the framework achieves 96.73% closed-set accuracy, 71.25% balanced open-set rejection accuracy, and strong clustering performance for unknown generators—ARI of 0.81, NMI of 0.90, and purity of 87.74%—with incremental learning further boosting final purity to 99.23%.

Abstract

Recent advances in generative Artificial Intelligence have made synthetic face images increasingly realistic, creating new challenges for multimedia forensics. Source attribution methods should not only identify the generator of an image when the source is known, but also handle samples produced by previously unseen models. However, most existing approaches address synthetic face attribution in a closed-set setting, where all possible generators are available during training. This assumption does not hold in real-world scenarios, where new generators continuously appear and rejected samples should be organized rather than simply discarded. In this work we propose a pipeline for open-set synthetic face source attribution that combines known generator classification, energy-based OOD rejection, and unknown generator discovery. A classifier is trained on known generators using frozen I-JEPA embeddings, while rejected samples are represented by combining projected I-JEPA features with Forensic Self-Descriptors and then clustered to discover groups of unknown generators. We also extend the discovery stage to an incremental scenario, where rejected samples arrive over time. Experiments on the WILD dataset show that the proposed method achieves 96.73% closed-set attribution accuracy. In the open-set setting, energy-based rejection reaches 71.25% balanced accuracy, while rejected samples are clustered into meaningful unknown-generator groups, obtaining an ARI of 0.81, an NMI of 0.90, and an overall clustering purity of 87.74%. In the incremental setting, the discovered generator space is progressively extended while maintaining a final purity of 99.23%. Cross-dataset experiments suggest that the pipeline can operate beyond the original dataset distribution, although post-processing remains challenging.

Problem

• open-set attribution
• synthetic face detection
• unknown generator discovery
• multimedia forensics
• out-of-distribution rejection

Innovation

• open-set attribution
• synthetic face detection
• energy-based OOD rejection
• incremental generator discovery
• I-JEPA embeddings