February 25, 2019

Slixmpp gets OMEMO support

TL;DR: Developers can already experiment with the slixmpp-omemo plugin.
Please give us feedback on the tracker or in the channel!

After almost a year since I started working on the OMEMO encryption mechanism support for Slixmpp, I am happy to finally announce a first release. I would like to get feedback. I am sure there are still plenty of things to improve, and so I encourage developers to bring out their inner vandal, break it and report their findings.

This library provides an interface to python-omemo.

You can find the code at https://lab.louiz.org/poezio/slixmpp-omemo.
Documentation is available in the README, and there is also an echo bot, with lots of comments.

Thanks to Syndace and Daniel for the help with the OMEMO implementation, and mathieui and Link Mauve for the help on Slixmpp and moral support.

Separate repository

As you may have noticed, this plugin is served via a separate repository. This is for licensing purposes. As much as I like GPL and copyleft, Slixmpp is licensed under the MIT license, and this is probably not going to change. Fortunately for Slixmpp this split should not last forever.

The python-omemo library that is used – developed by Syndace – is a complete reimplementation of the Signal Protocol unlike python-axolotl, which is a port of the original library implemented in Signal.

There are bits that prevent him from releasing his library under MIT at the moment, I am not entirely sure to grasp all the details but this is being worked on.

Why OMEMO?

There are still lots of things to be improved in the OMEMO specification.

I would personally like to see what is usually called Full Stanza Encryption added to the spec. Today, an OMEMO implementation will only encrypt the plaintext (<body/>) part of messages you send, and either leak everything else (e.g., chatstates, receipts, corrections, xhtml-im), or effectively disable them, for privacy-conscious implementations.

I would also like to drop Forward Secrecy, in the context of Instant Messaging. And I would like to have a better way to manage all these device keys, fortunately there are people working on this already.

Not having all these options (or having them, in the case of Forward Secrecy) heavily degrades user experience in my opinion, and that is my main concern.

Not having OMEMO though, is also not great either for user experience, many implementations nowadays provide it, and some even enable it by default. This makes it impossible for us Slixmpp users to communicate without having to ask the sender to turn it off first.

While I would prefer to see other alternatives, this library should help with the current situation, and we can go back to work on fixing the world.

What’s next?

Apart from the tons of bugs that I’ll have to fix in the following days/weeks, now that we have the foundations next step is to implement OMEMO in Poezio.

Any help is welcome!

EDIT 2019-03-02: A small but important precision on Full Stanza Encryption. I did write “what is usually called”, because it does not actually consist in encrypting the full stanza. The server still needs to see routing information, as well messages hints (e.g., <no-copy/> or <no-store/>) designed for the server rather than the final recipient. A better name might be “Arbitrary Extension Element Encryption” (thanks Flow.)

© Maxime “pep.” Buquet 2019. Licensed under CC-BY-SA 4.0 unless specified.

Powered by Hugo & Kiss.