Live data from Hacker News

Recovering redacted information from pixelated videos

positive.security

11–20 of 70 posts

Re: Recovering redacted information from pixelated videos

#11
As University libraries have moved online, one reads many poorly scanned journal articles. I often wonder about taking the time to clean them up. What replaces temporal information here is the same characters appearing over and over.

So of course I read this article hoping to learn about an off-the-shelf tool that would do a great job of scanned text reconstruction. Alas, the best candidates were "no code available."

Re: Recovering redacted information from pixelated videos

#13
post #4

Yet again mosaic blur's inferiority to opaque rectangle is demonstrated.

More generally if you want to redact pixels, then don't replace them with information that depend on those pixel values. It doesn't necessarily mean a black rectangle, but that's certainly simple and it works.

I’m slightly curious as to how, in these cases, someone came to decide that blurring was the right way to do it. It seems unlikely that they never thought of simply blacking/ blanking the text, but if so, then blurring must have seemed preferable.

My best guess is that, having seen blurring of faces (which is arguably OK when one merely wants to avoid casual attempts at identification, while retaining a ‘natural’ look), they assumed this was the proper way to do it in all cases.

Re: Recovering redacted information from pixelated videos

#14

Earlier quoted context omitted.

More generally if you want to redact pixels, then don't replace them with information that depend on those pixel values. It doesn't necessarily mean a black rectangle, but that's certainly simple and it works.

I’m slightly curious as to how, in these cases, someone came to decide that blurring was the right way to do it. It seems unlikely that they never thought of simply blacking/ blanking the text, but if so, then blurring must have seemed preferable. My best guess is that, having seen blurring of faces (which is arguably OK when one merely wants to avoid casual attempts at identification, while retaining a ‘natural’ loo…

I think it leaves the visual interest behind. Black bars are both boring and jarring. It's aesthetically displeasing on many levels.

Unfortunately, part of "leaving the visual interest behind" is precisely "pixels that depend on their real underlying values"....

Re: Recovering redacted information from pixelated videos

#15
post #3

The article references the famous face-depixellating PULSE GAN. Let's remember that PULSE also gave us Barry O'Bama. https://www.google.com/amp/s/www.theverge.com/platform/amp/2...

Indeed, however one has to note that this article is mainly about pixelated text, which is a bit different, especially considering that it (usually) has a known alphabet and you often even know what the individual font glyphs look like from context.

Re: Recovering redacted information from pixelated videos

#16

Earlier quoted context omitted.

More generally if you want to redact pixels, then don't replace them with information that depend on those pixel values. It doesn't necessarily mean a black rectangle, but that's certainly simple and it works.

I’m slightly curious as to how, in these cases, someone came to decide that blurring was the right way to do it. It seems unlikely that they never thought of simply blacking/ blanking the text, but if so, then blurring must have seemed preferable. My best guess is that, having seen blurring of faces (which is arguably OK when one merely wants to avoid casual attempts at identification, while retaining a ‘natural’ loo…

Retaining a "natural" look is phrased well. I think that's all behind the decision to infrasample (opposite of supersample?) parts of an image.

Re: Recovering redacted information from pixelated videos

#19

Even the best can make obvious mistakes. "I hacked a hardware crypto wallet and recovered $2M [video]" https://news.ycombinator.com/item?id=30067340 Showing a blurry 16 out of 24 trezor wallet seed words https://youtu.be/dT9y-KQbqi4?t=1720

This is somewhat bad. Trezor's dictionary is only 2048 words. So to brute force 8 remaining words would only take 2048^8 attempts. Which is 88 bits of security.
Post reply on HN