2026-09-27 · America/Los_Angeles · 论文 · #13

FARE: Forensic Acceptance Region Estimation for Catching Bait-and-Switch Image Generators

Modern AI image generators are increasingly deployed as opaque APIs, where customers can query the deployed service, but cannot inspect model weights or architecture. This creates a practical challenge: a provider may pass governance certification with one generator and later silently switch to a cheaper and lower-quality one for deployment, compromising public trust or even safety in high-stakes domains. We study integrity auditing at deployment time and propose FARE (Forensic Acceptance Region Estimation). A certified generator is enrolled by training FARE on images sampled from that…

热度 14.0 / 100；排名与评分保留该期记录。