☆ Yσɠƚԋσʂ ☆@lemmy.ml to Technology@lemmy.mlEnglish · 2 days agoChina Has Matched Anthropic in Cybersecurity, Resetting AI Racewww.wsj.comexternal-linkmessage-square49fedilinkarrow-up137arrow-down11file-textcross-posted to: technology@midwest.social
arrow-up136arrow-down1external-linkChina Has Matched Anthropic in Cybersecurity, Resetting AI Racewww.wsj.com☆ Yσɠƚԋσʂ ☆@lemmy.ml to Technology@lemmy.mlEnglish · 2 days agomessage-square49fedilinkfile-textcross-posted to: technology@midwest.social
minus-square☆ Yσɠƚԋσʂ ☆@lemmy.mlOPlinkfedilinkarrow-up3arrow-down1·4 hours agoThe context here is obviously software exploits given that we’re talking about LLM finding them.
minus-squareTrippinMallard@lemmy.mllinkfedilinkarrow-up1·edit-24 hours agoThat was not obvious to me. LLMs have been used for finding hardware, firmware, RF, software, and social exploits. RAM side-channel attacks are a good example of software exploits that are harder to exploit than find the vulnerability.
minus-square☆ Yσɠƚԋσʂ ☆@lemmy.mlOPlinkfedilinkarrow-up3arrow-down1·4 hours agoSure, you can do all that as well, but the context is an article about cyber security.
minus-squareTrippinMallard@lemmy.mllinkfedilinkarrow-up1·3 hours agoCybersecurity includes finding hardware, firmware, RF, software, and social exploits.
minus-square☆ Yσɠƚԋσʂ ☆@lemmy.mlOPlinkfedilinkarrow-up2·2 hours agoAgain, I’m not disagreeing that you can use LLMs to audit all these things. All I’m saying is that software is by far the easiest place to apply models and actually try out exploits end to end.
minus-squareTrippinMallard@lemmy.mllinkfedilinkarrow-up1·2 hours agoYour original comment was: Finding them is a prerequisite to exploiting them, and by far the hardest part. Once you know what the exploit is, abusing it is not difficult.
minus-square☆ Yσɠƚԋσʂ ☆@lemmy.mlOPlinkfedilinkarrow-up2·2 hours agoYes, and as I’ve repeatedly clarified here, I was making that statment in the context of software.
The context here is obviously software exploits given that we’re talking about LLM finding them.
That was not obvious to me. LLMs have been used for finding hardware, firmware, RF, software, and social exploits.
RAM side-channel attacks are a good example of software exploits that are harder to exploit than find the vulnerability.
Sure, you can do all that as well, but the context is an article about cyber security.
Cybersecurity includes finding hardware, firmware, RF, software, and social exploits.
Again, I’m not disagreeing that you can use LLMs to audit all these things. All I’m saying is that software is by far the easiest place to apply models and actually try out exploits end to end.
Your original comment was:
Yes, and as I’ve repeatedly clarified here, I was making that statment in the context of software.