• 0 Posts
  • 13 Comments
Joined 5 months ago
cake
Cake day: June 12th, 2025

help-circle






  • bluemoon@piefed.socialtoPrivacy@lemmy.mlA list of privacy tips
    link
    fedilink
    English
    arrow-up
    1
    ·
    edit-2
    9 hours ago

    i am reminded of the Linux.Conf.Au 2017 seminar “how to disappear” (iirc) which lists all the then existant ways of tripping up surveillance (physical device • AI processing • identifyability of your personhood) involving atleast reflex-trinkets to disturb surveillance cameras especially at night, secondly wearing stickers that trip up AI models and inducing hallucination as well as shining blue light on points of your face to appear as a different person entirely to the AI model, thirdly either covering specifically the nose & eyes since these contained the most identifiable points in a face (according to seminar speaker’s presentation.) just recollected off the top of my head.

    also dot-matrix writers to ward off printer software.

    also also remined by my converting image files to plaintext formats like .xpm .xpg .xpb .ppm .ppg .ppb et al. that strips away all meta-data in the image file from my testing. then altering it. then converting that altered plaintext image file to whatever desired fileformat.

    as for recordings of the space you’re in: i recall that noise-jammer that obfuscates AI models fed the recordings. just a background noise to play. i forget the channel that presented this, professional musician with a company that also develops anti-AI treatment to audio files like nightshade or glaze but actually unstrippable. he showcased the anti-AI audio jammer.

    as for actual listening in… blankets covering both of your faces, loud music in background that plays at the same tones you’re speaking in, then your company whispers?

    theoretically: do clothes with metal fibres woven in (faraday cage style) hinder Pulse-Fi & WhoFi ? is there a way to trip up detection by appearing as many heartbeats & a mass of people? that same conference talk brought up textile pattern with facial prints that get recognized by AI models as additional people.

    truly appreciate the post

    also time to put gradients on images with text like it’s a pre-Y2K slide