Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Why blurring sensitive information is a bad idea (2007)

dheera.net

Post reply on HN
Print view View original

51–60 of 131 posts

Re: Why blurring sensitive information is a bad idea (2007)

#51

Post by ColinDabritz »

nathell wrote: post #30

In some cases, coloring over parts of the image might still not be enough. Specifically, when all of the following are true: (i) the domain of possible entries is reasonably small (e.g., a number or a name and surname), (ii) the text is printed in a proportional (not fixed-width) typeface, and (iii) enough of the rest of the line is visible to infer the font size and kerning settings.

I think this form of analysis is amazing. It breaks the "The best solution was to actually cut out the parts of the page you don't want seen and then scan in the result." approach from elsewhere in this thread as well. Even cutting out each portion of each line isn't enough because you can reconstruct line sizes. You could cut out each individual word and string them together, with a single size "redaction" block perhaps, but that's a lot of work.

Re: Why blurring sensitive information is a bad idea (2007)

#52

Post by bryondowd »

userbinator wrote: post #38
nathell wrote: post #30

In some cases, coloring over parts of the image might still not be enough. Specifically, when all of the following are true: (i) the domain of possible entries is reasonably small (e.g., a number or a name and surname), (ii) the text is printed in a proportional (not fixed-width) typeface, and (iii) enough of the rest of the line is visible to infer the font size and kerning settings.

Actually, using fixed-width would be even easier since it reveals the length.

I'd be curious to see some numbers backing one or the other, but I suspect proportional would be better. Fixed width narrows to a number of characters. So it depends how many possible values have x characters. Proportional narrows to a particular width, which could be more granular than character count. So it depends how many possible values turn out to be exactly x pixels across given the type settings.

Re: Why blurring sensitive information is a bad idea (2007)

#53

Post by DanBC »

hawleyal wrote: post #33

Not really. You would need to have a copy of an uncovered version to know how many lines of text, font size, kerning et al.

Some documents are trivially easy to get near copies.

Other documents provide their own context - they have a lot of other text - that you can use to get font, font size and spacing and etc.

Re: Why blurring sensitive information is a bad idea (2007)

#54

Post by frogpelt »

danepowell wrote: post #27
bithush wrote: post #22

I feel this is also very relevant http://en.wikipedia.org/wiki/Christopher_Paul_Neil Police took a photo with a "swirl" effect of a paedophiles face and reversed it to reveal a very usable picture. So good in fact he was found and arrested.

He wasn't just a pedophile, he was a child molester.

"just a pedophile"

A phrase I never thought I'd see.

Re: Why blurring sensitive information is a bad idea (2007)

#55

Post by freehunter »

Houshalter wrote: post #43
MichaelApproved wrote: post #31

Earlier quoted context omitted.

You don't have to go all the way back to the early 2000s to find poorly redacted federal documents. Rachel Maddow had a great segment on botched redactions, like removing the black lines from a PDF. The best solution was to actually cut out the parts of the page you don't want seen and then scan in the result. All other methods are prone to mistakes. Edit: corrected link to segment http://www.msnbc.com/rachel-maddow-…

Would a screenshot work?

I've always drawn over sensitive parts in Paint or something, then taken a screenshot of it. It makes sure that everything is compressed to a single layer and can't be undone.

Re: Why blurring sensitive information is a bad idea (2007)

#56

Post by pests »

Houshalter wrote: post #44
jamuraa wrote: post #32

Earlier quoted context omitted.

You can in fact do this, it is a technique known as Super-Resolution and has been around for at least a decade. http://ieeexplore.ieee.org/xpl/articleDetails.jsp?arnumber=5...

That's really cool. I wonder if that could be used to get images of other planets.

gus_massa posted a link about Pluto:

http://www.boulder.swri.edu/~buie/pluto/mapstory.html

Re: Why blurring sensitive information is a bad idea (2007)

#57

Post by exelius »

jebus989 wrote: post #37
bithush wrote: post #22

I feel this is also very relevant http://en.wikipedia.org/wiki/Christopher_Paul_Neil Police took a photo with a "swirl" effect of a paedophiles face and reversed it to reveal a very usable picture. So good in fact he was found and arrested.

Police == 4chan?

Yeah, basically. 4chan's admins have been working with the police for many years because of the questionable shit that gets posted on /b/ (child porn, murder threats/evidence, suicidal posts, etc.) They would have been shut down years ago if they weren't cooperating.

Re: Why blurring sensitive information is a bad idea (2007)

#58

Post by jebus989 »

exelius wrote: post #57
jebus989 wrote: post #37

Earlier quoted context omitted.

Police == 4chan?

Yeah, basically. 4chan's admins have been working with the police for many years because of the questionable shit that gets posted on /b/ (child porn, murder threats/evidence, suicidal posts, etc.) They would have been shut down years ago if they weren't cooperating.

The way I've heard this story in the past was specifically that anon did the unswirl and then tipped off the police. But seeing as that isn't mentioned on wiki (and not uber keen to google about), I guess it's probably not the case.

Re: Why blurring sensitive information is a bad idea (2007)

#59

Post by Udo »

ekr wrote: post #5
Udo wrote: post #2

A nice Gaussian blur would probably be fine, it's specifically the pixelation technique that's leaking data.

You think the technique mentioned in the article doesn't leak data? It actually leaks more data than a gaussian blur. Gaussian blur suffers from exactly the same problem, although a different difference function is needed for it.

I see that I deserved a schooling on gaussian blur, but

  > You think the technique mentioned in the article doesn't leak data?
yes, I do think it leaks data, which was the correct half of my comment, actually.

Not that it matters now, but I always assumed blurring distorted the data in way way that would make it hard to recover in practice.

Re: Why blurring sensitive information is a bad idea (2007)

#60

Post by Udo »

jarman wrote: post #11
Udo wrote: post #2

A nice Gaussian blur would probably be fine, it's specifically the pixelation technique that's leaking data.

It will be worse than that. Pixellation destroys information, it's not reversible in general. But in this case, what is left is enough for recovery, as source domain (check number) is small. Gaussian blur is theoretically lossless (rounding loss and loss on image edges in practice) - you can take arbitrary image, blur it, reverse and get original+noise

Thanks for making me look deeper into what's going on with common blurring algorithms. My instincts were clearly wrong here, for years I thought reversing blurs wasn't feasible in practice. I was very, very wrong.
Post reply on HN