Re: [libreplanet-discuss] HTML tags that should be stripped for privacy reasons.

[Josh Branning]

On 28/01/17 03:31, Mike Gerwitz wrote:
> On Fri, Jan 27, 2017 at 13:29:52 +0000, Josh Branning wrote:
> > What I would really like is implementation details on how to avoid many of
> > the issues stated on this page:
> > https://en.wikipedia.org/wiki/Internet_privacy
> >
> > So like two columns; Stated problem, and possible solution.
>
> This isn't a problem that can be summarized as such with any meaningful
> level of detail.
>
> I'm not entirely sure what you mean by handling it on the
> webserver.  Is this like a privacy-respecting proxy server (in which
> case you can look into things like Privoxy[0])?  The other concern there
> is that you'd have to MitM HTTPS connections (with the user's consent,
> of course), which is dangerous.

The web server part will ideally only be an interface (like a HTTP 
proxy) to the network.

MitM attacks are mitigated in a way similar to Convergence. [1]

[1] https://en.wikipedia.org/wiki/Convergence_(SSL)

> There are many things that you'd have difficulty detecting outside of
> the user's browser---you'd be doing static analysis on behaviors that
> aren't entirely defined until runtime, or might be difficult to
> detect.  If you don't entirely strip out JS, it's not possible to
> mitigate most things unless you execute the code on your server and
> analyze it (since it can just modify the DOM).
>
> Ultimately, the user can trust only their web browser, not a remote
> server, so mitigations also have to be done there.  And the level of
> complexity there is staggering (see NoScript; Privacy Badger; uBlock
> Origin; etc).

Accessing third-party web servers from the browser and expecting 
anonymity or privacy, I agree, is not the way forward. The server is to 
be run on localhost by the user of the software.

> With regards to information, the Tor Browser has documented some
> implementation details[1].
>
> One of the best ways to learn how to mitigate issues is to know how to
> exploit them, in practice and in theory.  OWASP has some good material
> (duplicated in various projects, granted) that's readily available.[2]
>
> You can also look at commits in repositories of these various projects.
>
>
> [0]: http://www.privoxy.org/
> [1]: https://www.torproject.org/projects/torbrowser/design/
> [2]: https://www.owasp.org/index.php/Category:OWASP_Project

Thank you for these links, they have given me much to think about.