← feed
NOTABLE
arXiv: SURA reuses one small clean proxy to keep unlearning-based backdoor detection working in federated learning
arXiv 2611.13491 proposes SURA, a data-efficient unlearning-based backdoor detector for federated learning: it probes client models with train-mode loss ascent reusing the same small clean proxy across steps, and restores batch-norm running statistics so repeated probing does not confound the measurement of malicious clients.
released 2026-11-10 · ingested 2026-11-11T12:15Z
Sources
Primary source
Cite this
APA
AI Security Index. (2026). arXiv: SURA reuses one small clean proxy to keep unlearning-based backdoor detection working in federated learning. AI Security Index. https://arxiv.org/abs/2611.13491
BibTeX
@misc{aisec-2026-11-11-arxiv-2611-13491-sura-unlearning-backdoor-detection-federated-learning,
title = {arXiv: SURA reuses one small clean proxy to keep unlearning-based backdoor detection working in federated learning},
year = {2026},
howpublished = {\url{https://arxiv.org/abs/2611.13491}},
note = {AI Security Index item 2026-11-11-arxiv-2611-13491-sura-unlearning-backdoor-detection-federated-learning}
}