Flameshot: Feature: Mosaic

Created on 5 Jun 2018  路  4Comments  路  Source: flameshot-org/flameshot

Flameshot's Blur isn't very good, as pointed out in #201 and #121.

image

As an alternative, I would suggest a mosaic option instead.

image

Even a pixel size of 4 is enough!

Most helpful comment

While it is not easily human readable, I'm not convinced it's impossible to recover the original text, especially if you can make some assumption about the text and font and know the algorithm (see [1]).
In the example, even with the bare eye it is possible to make out word boundaries and character heights, which might be enough to guess the content.
The required pixel size is also obviously depending on the font size, so it would have to be configurable or auto detected.

I think the blur tool should display a warning against using it on sensitive information and to use the box tool instead.

A better implementation with pixelation/mosaic or a combination of them would be nice nevertheless.
Maybe there even is a way to make a secure blurring algorithm. Something like detecting text and replacing it with bogus text before blurring. That's the way I would do it manually if I wanted to be secure and maintain the overall looks.

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

All 4 comments

While it is not easily human readable, I'm not convinced it's impossible to recover the original text, especially if you can make some assumption about the text and font and know the algorithm (see [1]).
In the example, even with the bare eye it is possible to make out word boundaries and character heights, which might be enough to guess the content.
The required pixel size is also obviously depending on the font size, so it would have to be configurable or auto detected.

I think the blur tool should display a warning against using it on sensitive information and to use the box tool instead.

A better implementation with pixelation/mosaic or a combination of them would be nice nevertheless.
Maybe there even is a way to make a secure blurring algorithm. Something like detecting text and replacing it with bogus text before blurring. That's the way I would do it manually if I wanted to be secure and maintain the overall looks.

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

You make a very interesting point. I looked over that link and I do agree with you - a good solution to the problem of anonymizing an image is therefore to warn about using blur/mosaic and recommend box instead.

I do still think that mosaic is a nice option to have, though.

Mosaic is nice for maintaining context and anonymizing without obliterating the fact that there was information there before. Pixelation and blurring can already be undone to some extent by AI methods, so you're also right that a box is the only true redaction. Perhaps a random or "pseudotext" mosaic that just samples the colors in the space would do the job.

This is merged to master as the new blur tool.

Was this page helpful?
0 / 5 - 0 ratings

Related issues

dajare picture dajare  路  4Comments

hosiet picture hosiet  路  4Comments

borgmanJeremy picture borgmanJeremy  路  3Comments

thiswillbeyourgithub picture thiswillbeyourgithub  路  3Comments

ronjouch picture ronjouch  路  3Comments