Researchers develop hardware defenses for AI chip bit-flip attacks

A team from the University of Florida, UT Dallas, and Rensselaer Polytechnic has designed a system called BARE-AI that detects and counters bit-flip attacks on AI chips in real time. These attacks can quietly alter just a few memory bits and throw off critical AI decisions. The new approach embeds monitoring directly into the chip, aiming to keep neural network accuracy intact during attacks.

• New system detects and mitigates AI chip bit-flip attacks
• BARE-AI monitors chip performance in real time
• Targets deep neural networks used in safety-critical systems
• Doesn't require extra training or major hardware changes