Researchers say they embedded malware in up to half of the neurons in the benchmark AlexNet AI model⁠ and still kept the model's accuracy rate above 93.1% (Radhamely De Leon/VICE)

Radhamely De Leon / VICE:
Researchers say they embedded malware in up to half of the neurons in the benchmark AlexNet AI model⁠ and still kept the model's accuracy rate above 93.1%  —  In a proof-of-concept, researchers reported they could embed malware in up to half of an AI model's nodes and still obtain very high accuracy.

from Techmeme https://ift.tt/36XcpxE
https://ift.tt/3hZCX7K July 24, 2021 at 03:15AM https://ift.tt/gaIxo5