As someone who’s been dealing with a lot of low mood lately, it really brightens my day to hear someone express gratitude for my anonymous, unauthorized trail maintenance.

Text blurred for privacy reasons.

    • letsgo@lemm.ee
      link
      fedilink
      English
      arrow-up
      4
      ·
      1 day ago

      There’s quite a lot of information in those pixels. You can make out gaps between words, risers and descenders, two smileys in the first paragraph which I reckon are probably a smiley and a thumbs up which will further narrow down the possibilities for the words in their vicinity. Constrain that further with English language rules and I reckon that’d get you most of the way there.

      • okamiueru@lemmy.world
        link
        fedilink
        arrow-up
        4
        ·
        1 day ago

        Not to mention that you know that exact typeface and pixel perfect location where letters can be, so it should be relatively easy to go through each possible subsequent character and match the pixelated value.

      • Duamerthrax@lemmy.world
        link
        fedilink
        English
        arrow-up
        8
        ·
        2 days ago

        The example Unredacted used is 4 pixels high. The one in the screenshot seems to only be 2 pixels high.

        • Elvith Ma'for@feddit.org
          link
          fedilink
          arrow-up
          6
          ·
          2 days ago

          True, this would make it harder. But… On the other hand, its not a random password but text. If you know (or guess) the language you may be able to employ other tricks like “how common is each letter?”, “which combination of letters is more common in this language?” And so on. Maybe the hidden markov model mentioned in the research paper does that (which would be one thing that Markov Models do IIRC).

          • CaptainBlagbird@lemmy.world
            link
            fedilink
            arrow-up
            3
            ·
            2 days ago

            Right, technically even one pixel per letter could be solvable. Different letters would mostly result in a slightly different hue. And if multiple letters have the same, it could be guessed via the neighbours and statistical frequency of each letter.

            We also have the context and could specifically look for words like trees, path, thank, saw first.

            • Duamerthrax@lemmy.world
              link
              fedilink
              English
              arrow-up
              1
              ·
              1 day ago

              Isn’t this software available? People keep telling me that it could be decensored, but no one has tried it.

      • tigeruppercut@lemmy.zip
        link
        fedilink
        arrow-up
        2
        ·
        2 days ago

        I wonder if an entire line gets blurred with one pattern-- therefore if you know what the emojis are supposed to look like it can help figure out the blurring pattern