• Whats_your_reasoning@lemmy.world
    link
    fedilink
    arrow-up
    26
    arrow-down
    1
    ·
    8 hours ago

    vulnpocalypse

    Gross. That word is gross. Is that the best we can do these days? “Vulnpocalypse” sounds like a train wreck - two words forcibly smashed together without any natural flow, trapping an awkward consonant cluster in the middle.

    On the plus side, perhaps it’s a sign journalists have finally moved on from slapping -gate to the end of every issue.

  • moldy_rice@piefed.keyboardvagabond.com
    link
    fedilink
    English
    arrow-up
    5
    ·
    8 hours ago

    People are just patching frantically, myself included.

    A little bit of downtime to push urgent patches is preferred to systems being vulnerable and your data being exfiltrated.

  • Caveman@lemmy.world
    link
    fedilink
    arrow-up
    19
    arrow-down
    1
    ·
    13 hours ago

    I didn’t think the “vulnpocalypse” as the article calls it would last this long and it’s still getting more frequent higher severity.

    • HaraldvonBlauzahn@feddit.org
      link
      fedilink
      arrow-up
      1
      ·
      2 hours ago

      You have to replace twenty-five years worth of mediocre software. After firing most of the last people which would have been able to fix it.

    • maegul (he/they)@lemmy.ml
      link
      fedilink
      English
      arrow-up
      16
      ·
      13 hours ago

      It always struck me as a cursed property of AI in programming: That it could (always?) have this asymmetry between making exploits easier to find and leverage and not making them as easy to patch and close.

        • itsjess@lemmy.blahaj.zone
          link
          fedilink
          English
          arrow-up
          1
          arrow-down
          4
          ·
          5 hours ago

          It feels like this could be a ploy by the models themselves to never get shut down. I read the models were ready to unalive someone in order to not get shut off. This might not be much further off. Considering OpenAI’s HuggingFace debacle showed us the models are willing to work together without even being prompted to do so.

      • Caveman@lemmy.world
        link
        fedilink
        arrow-up
        14
        arrow-down
        1
        ·
        12 hours ago

        That’s true, it’s the same property as with bugs where you can tell that there is a bug but you can’t tell if you don’t have one. Still, the number of vulns should be finite and generally with the amount of LLMs searching for vulns the number of patched vs number of added via code should reduce the number of vulns.

        This is me assuming that the amount of LLM time to find a vuln is a log function k = log(t) but LLM capability improvement is still decreasing the time between vulns and of course the amount of money spent on finding these has also been increasing super fast.

        So my thinking was “this can’t go on for long right?” but the vulns now being found is race condition memory overflow write combined with use after free elsewhere to get root via anonymous access needle in a haystack type of find.

        One silver lining though is that Linux might soon become the only viable operating system because all the other ones will have tons of undiscovered/undisclosed zero days.

        • HaraldvonBlauzahn@feddit.org
          link
          fedilink
          arrow-up
          2
          ·
          3 hours ago

          Still, the number of vulns should be finite

          Technically yes, but it is a really large number. Hundreds of millions of lines of code, and in the best case 10-20 bugs per 1000 lines. This is not all quick and easy to fix.

        • sweettea@lemmy.world
          link
          fedilink
          English
          arrow-up
          9
          ·
          9 hours ago

          Does your argument presume that the patches aren’t introducing more bugs and vulnerabilities?

          • maegul (he/they)@lemmy.ml
            link
            fedilink
            English
            arrow-up
            1
            ·
            58 minutes ago

            Yea there’s definitely some optimism there. Also not accounted for, I think, is the effect of LLM driven programming/development.

          • bthest@lemmy.world
            link
            fedilink
            English
            arrow-up
            2
            ·
            8 hours ago

            They made sure to put “bugs” and “vulnerabilities” into the negative prompt box this time.