News for hackers. Stuff that matters.

A comment on Adversarial image attacks are no joke. Parent comment

Re:Adversarial image attacks are no joke by pixelgeek on Monday November 29, 2021 @11:27AM (#29381709)
But that is not a vision issue. That is providing people with incorrect information.
[ Parent | on HN ]
Re:Adversarial image attacks are no joke by b3morales on Monday November 29, 2021 @12:20PM (#29382424)
The specific trick doesn't really matter; the point is that it's possible to maliciously create a situation that makes human pilots act dangerously. We accept that the possibility can't be made nil, and we have post facto rules to deal with it. The same principle applies to traps for machines.
[ Parent | on HN ]
Re:Adversarial image attacks are no joke by ClumsyPilot on Monday November 29, 2021 @03:21PM (#29384734)
Nope, it 0.1% of humans crash but 100% of teslas crash that's not 'the same'
[ Parent | on HN ]
Re:Adversarial image attacks are no joke by MauranKilom on Monday November 29, 2021 @07:35PM (#29387298)
So would it be better/fine if Tesla randomized the CV ML models to some extent? I actually feel the answer is probably "yes", surprisingly...
[ Parent | on HN ]
Re:Adversarial image attacks are no joke by qw501428 on Monday November 29, 2021 @11:36AM (#29381838)
Perhaps a mirror/reflective sticker that blinds drivers near a sharp curve?
[ Parent | on HN ]