[GNU/consensus] PGP as web standard

[Guido Witmond]

Hello!
So we need the software to be received from trusted source. Just like we
receive browser from the repositories of our GNU/Linux system, which are
trusted and have some ways to prove the software was not modified.

Indeed.

We could use browser extensions to implement end-to-end encryption, but
they hardly can be convenient, because they will always lack some
important information, for example information on your recipient, that
could be received only with the help of the website. This information is
required to implement such useful things as automatic key selection for
the recipient.

Indeed.

So the problem of end-to-end encryption in web-based software requires
that browser interacts with the website receiving required information,
but all the software that does the encryption loads from the computer of
user, not from the Internet at web site load time.

Indeed.

All that led me to the idea, that implementation of end-to-end
encryption in web technologies should be a part of web standard, so the
rules of interactions between website and browser encryption module is
defined strictly. For example, it could be defined as special kind of
forms, that is filled by user with unencrypted text, but when you submit
the form, browser really sends it PGP encrypted using key, that is
determined by the context (you are in a web chat with certain
recipients). This also can help in providing some extra security for
this text blocks, because browser could isolate unencrypted text from
any javascript, that may want to read it.

That's what I've designed and prototyped, except for PGP and Javascript
:-) I avoid all the accumulated cruft of those programs and implement it
on top of TLS and private CAs, one for each website.

So, here is my question to you, as a social networking project members.
Have you ever though of necessity of realizing end-to-end encryption as
part of web standard? Do you think it is possible to push? Maybe it is
nevertheless possible to implement end-to-end encryption with some
javascript using some extra security and isolation measures? Or maybe
you have some other ideas how to implement it, that I didn't think of?

Take a look at Eccentric Authentication.

Please read:

http://eccentric-authentication.org/blog/2012/10/23/a-blog-site.html
http://eccentric-authentication.org/blog/2013/06/07/run-it-yourself.html

Or read :

http://eccentric-authentication.org/blog/2014/11/30/spot-the-differences.html
and: http://eccentric-authentication.org/Usable-Security.pdf

With regards, Guido Witmond.
eccentric-authentication.org

[senya]

Post by Guido Witmond
http://eccentric-authentication.org/blog/2012/10/23/a-blog-site.html
http://eccentric-authentication.org/blog/2013/06/07/run-it-yourself.html

http://eccentric-authentication.org/blog/2014/11/30/spot-the-differences.html

Post by Guido Witmond
and: http://eccentric-authentication.org/Usable-Security.pdf
With regards, Guido Witmond.
eccentric-authentication.org

Thanks, great job!

I'll look at it closer. It would be great to integrate it with
diaspora*, however it for sure requires much work. But it is a
definitely the right way to do things.

[carlo von lynX]

Have you ever though of necessity of realizing end-to-end encryption as
part of web standard? Do you think it is possible to push? Maybe it is

It has been done. Web Crypto API is being deployed. It is a terrible
idea that Eleanor Saitta and I have heavily criticized in the
respective W3C mailing lists because there is no way on earth that
servers can be trusted to deliver the correct HTML and Javascript
such that end-to-end encryption will actually take place - therefore
it is giving false promises to users. There is no privacy gain
compared to using HTTPS and having the server handle the data
since users would never be able to tell when their trust is broken.

nevertheless possible to implement end-to-end encryption with some
javascript using some extra security and isolation measures? Or maybe
you have some other ideas how to implement it, that I didn't think of?

Since the web browser by definition shows what the server tells it
to show, the user cannot tell by looking at the web's user interface
if it is doing crypto in the browser or not.

The web is only safe when the server is "localhost" - and even that
is hard to hammer into user's heads.

[senya]

Post by carlo von lynX
It has been done. Web Crypto API is being deployed. It is a terrible
idea that Eleanor Saitta and I have heavily criticized in the
respective W3C mailing lists because there is no way on earth that
servers can be trusted to deliver the correct HTML and Javascript
such that end-to-end encryption will actually take place - therefore
it is giving false promises to users. There is no privacy gain
compared to using HTTPS and having the server handle the data
since users would never be able to tell when their trust is broken.

This is a good explanation why javascript solutions don't do. Thanks.