Think your proprietary AI is safe behind closed doors? Xiao et al. introduce ModelSpy, a side-channel attack that steals complete neural network architectures from up to six meters away just by listening to a GPU's electromagnetic leakage through a wall. ndss-symposium.org/ndss-paper/

0

If you have a fediverse account, you can quote this note from your own instance. Search https://techpolicy.social/users/joebeone/statuses/116184180548062850 on your instance and quote it. (Note that quoting is not supported in Mastodon.)