Hacker News new | ask | show | jobs
by zamalek 2244 days ago
This is a great idea! You might reconsider using blur to obscure the information, as it is trivial to defeat [1] (no ML required!).

[1]: http://dheera.net/projects/blur

2 comments

True, blur on large areas is reversible, even visually user can see the text true and I hope that users will recognise that. That's why I also added scramble option which replaces all text with random characters. Combining scramble for data protection and blur for visual affect is the best combination
It would be cool to automatically scramble the text behind the blur. This would give the look without leaving the original data there.
That's a good idea. I could add new secure/maximum privacy action in the extension UI. Instead of relying on users knowing about the limitations of blur
Security shouldn't be opt-in; you should be secure by default because most users will stay with the default (which is reasonable - the entire point of the extension is to prevent information leaks, and as the author you're the domain expert and your users should be able to trust your judgement which comes in the form of a default).

Also, may I propose that as replacement text before blurring you use ads targeted at computer graphics researchers. /jk

How about two blurs? I've thought that a rotational blur followed by a gaussian or similar would be pretty secure.
"pretty secure"

It's either completely secure or it isn't secure at all.

Blurs can be deconvolved if the point spread function is known or can be guessed. Two blurs is just security through obscurity and hoping that numerical precision / stability issues work out in your favor.

Just black out sensitive areas.

https://en.m.wikipedia.org/wiki/Deconvolution

http://refocus-it.sourceforge.net/

That's an example of bruteforce, but the blur itself can actually be reversed sometimes: http://yuzhikov.com/articles/BlurredImagesRestoration1.htm