SPRINT: Secret Pixel Reconstruction for Robust AI Image Attribution
Researchers have introduced SPRINT (Secret Pixel Reconstruction fingerprinting), a novel method for attributing AI-generated images to their source model. Unlike existing fingerprinting techniques that rely on publicly discoverable patterns, SPRINT uses a secret to define hidden reconstruction targets, making the verification task private. This design provides robustness against adaptive attacks, where knowledge of the fingerprinting technique can be exploited to perturb fingerprints and evade detection. The method addresses a growing accountability problem in detecting the source model of AI-generated images, as current fingerprints are extremely brittle under adaptive attacks. SPRINT keeps the verification task itself private, preventing attackers from seeing the task the verifier solves. The work was published on arXiv with ID 2508.05691.
Key facts
- SPRINT is a model attribution method for AI-generated images.
- It relies on a secret to define hidden reconstruction targets.
- The method provides robustness to adaptive attacks.
- Existing fingerprinting techniques are brittle under adaptive attacks.
- SPRINT keeps the verification task private from attackers.
- The work addresses the accountability problem of AI image source detection.
- The paper is available on arXiv with ID 2508.05691.
- The announcement type is replace-cross.
Entities
Institutions
- arXiv