• LainTrain@lemmy.dbzer0.com
    link
    fedilink
    English
    arrow-up
    24
    ·
    4 months ago

    What I don’t understand is why the TTS key could even delete voices or read past responses from other devices, ideally each device should have its own properly scoped API key that only lets it access the immediately necessary functionality and no more.

    • ReveredOxygen
      link
      fedilink
      English
      arrow-up
      3
      ·
      4 months ago

      I would imagine that the devices aren’t making elevenlabs requests directly, but just making requests to the rabbit backend, which forwards the responses. if I’m wrong, then that’s quite impressively bad security

      • LainTrain@lemmy.dbzer0.com
        link
        fedilink
        English
        arrow-up
        2
        ·
        edit-2
        4 months ago

        Then wouldn’t it be just one API key to the rabbit backend instead? The researchers are suggesting it’s several keys though. Or are you suggesting every device has the same key to Elvenlabs that it sends over to the rabbit backend which passes that through to the request? That’s also very silly if they did that.

        • ReveredOxygen
          link
          fedilink
          English
          arrow-up
          2
          ·
          4 months ago

          My understanding was that they leaked the key that the rabbit backend uses to make requests to elevenlabs, and were just too lazy to change it. I could easily be wrong though

          • LainTrain@lemmy.dbzer0.com
            link
            fedilink
            English
            arrow-up
            1
            ·
            4 months ago

            I don’t think that’s the case, because otherwise how did they leak this key that the backend uses, that presumably stayed in the backend, by reverse-engineering the rabbit android application?

            I think the devices all just have hardcoded keys to the APIs themselves.