4 comments

[ 2.0 ms ] story [ 18.4 ms ] thread
"An AI-based malware blocker with no whitelist"

What could possibly go wrong?

Apearantly MS thought that nothing would :) genius.. If only they had the foresight not release this type of stuff before running proper QA
Things like this: https://www.bleepingcomputer.com/news/security/microsoft-def...

Maybe that is a good idea to remove bloat from windows computers

Exactly.

It seems like it has a lot of potential for things to go wrong. I can imagine the system flagging all sorts of valid software as malware for various reasons and failing to flag actual malware that does not meet whatever inscrutable criteria the ML model uses.

Systems like this seem like a good idea on paper, but without some basic controls like a whitelist / safelist it invites all sorts of trouble.