BARE-AI uses built-in monitors to detect bit-flip attacks in AI accelerators

Researchers at the University of Florida, UT Dallas and RPI published BARE-AI, a runtime framework that detects, localizes and mitigates bit-flip attacks on deep neural networks during inference using built-in performance monitors, without retraining.

• BARE-AI operates at inference time and requires no model retraining
• The framework detects, localizes and mitigates bit-flip attacks
• It relies on built-in performance monitors in AI hardware
• Preprint arXiv:2610.08739 was published in October 2026