• flambonkscious@sh.itjust.works
    link
    fedilink
    English
    arrow-up
    4
    ·
    4 months ago

    …sorta. The complexity here is their driver is signed, but it’s also loading code from their channel file (that was all zeroed out), and it seems the necessary error checking wasn’t implemented.

    I haven’t yet got to the root cause they published, this is just what I gathered from the video of a retired MS kernel dev who posts stuff.

    Obviously with their design it allowed them to be flexible at the cost of playing with fire - I’m impressed they got away with it for so long, really

    • skuzz@discuss.tchncs.de
      link
      fedilink
      arrow-up
      2
      ·
      4 months ago

      Thank you for the clarification. WHQL is such a pain to set up, I’m sure the AV vendors whined, “but, security! Do we have to test everything every time? That would slow an urgent 0day release!”

      • flambonkscious@sh.itjust.works
        link
        fedilink
        English
        arrow-up
        1
        ·
        4 months ago

        Yeah, there’s some limits to what they could do while maintaining pace for the 0 day stuff…

        Some input validations would be the most basic things they should have done years ago. I’m aware of the hashing mature vendors do of any content they download for updates or deployments. Signature checking as well, and that’s before the code is even inspected - why don’t they include their automated tests they obviously aren’t using in the update as a sanity check client-side? (I’m not aware of anyone doing this or even if it’s possible without the rest of the IDE, stack, I’m no dev)