• Lemmilicious@feddit.nu
      link
      fedilink
      English
      arrow-up
      6
      arrow-down
      3
      ·
      2 days ago

      I know AI has been marketed as some big boost in finding vulnerabilities, but as far as I know that’s still all marketing talk without real substance. The important thing, using AI or not, is that finding vulnerabilities and fixing them requires intent, which seems to be the missing piece in this case.

      • Akuchimoya@startrek.website
        link
        fedilink
        English
        arrow-up
        5
        ·
        1 day ago

        It turns out it’s much easier to destroy things than to build things well. I can destroy (greatly destabilize and render unsafe) a building by renting a sufficiently large vehicle and ploughing it through the building. I have literally zero skills it takes to design and/or build a structurally sound building. Finding vulnerabilities in code is orders of magnitude different than building secure code in the first place.

      • filcuk@feddit.uk
        link
        fedilink
        English
        arrow-up
        8
        ·
        1 day ago

        Have you missed the practically weekly news about hundreds of old CVEs found across various established codebases?
        It’s just a matter of which company wants to flex their new model as an advertisement.

        • Lemmilicious@feddit.nu
          link
          fedilink
          English
          arrow-up
          1
          ·
          1 day ago

          Sorry for being obtuse, but I actually don’t understand how your comment relates to mine. But it does indeed seem to be a matter of which company wants to flex their model, like you say!

          • filcuk@feddit.uk
            link
            fedilink
            English
            arrow-up
            2
            ·
            1 day ago

            I was pointing out that AI has been making waves in finding vulnerabilities, contrary to you saying it appears to be nothing but marketing.

            It is a double-edged sword. Anyone with an agent and a few hundred bucks can find a weakness to exploit somewhere. Scammers are making fully automated exploits that are easier than ever to fall for. Absolute chaos.

            • Lemmilicious@feddit.nu
              link
              fedilink
              English
              arrow-up
              1
              ·
              1 day ago

              Ah yeah, then I get you, thanks! I’ve noticed those waves of marketing for sure, I haven’t looked into most of them but I was a bit burned by the Claude Opus crap that was very little bite in proportion to the bark. From what I read it mostly found the same vulnerability lots of times in different places, and seemingly not because it was an unusually hard to detect vulnerability, but rather because they took the time and effort to direct it in the right place.

              This is sort of what I mean, whether AI or not, it takes intent. Whether you think humans or AI do it better, in the success stories I’ve heard intent has been the common factor. That’s my main idea!