How can I get E2EE (end to end encryption) in a group meeting?

29 Nov 2024 · 2 min

Watch on YouTube

1:59

Short technical Q&A from a YouTube video by Tsahi Levent-Levi (recorded/uploaded Nov 29, 2024) about how to get end-to-end encryption (E2EE) in group meetings. Speakers explain that modern browsers expose insertable streams (aka MediaStreams Encoded Transform) which let applications access encoded media frames before WebRTC's SRTP encryption so apps can apply their own E2EE. They note encryption adds modest overhead (~32 bytes/frame for video), recommend using existing group-key management protocols (e.g., MLS or Matrix's approaches) rather than inventing custom crypto, and advise leaving key management to the application layer.

Questions this video answers

November 29, 2024.

Insertable streams (aka MediaStreams Encoded Transform).

About 32 bytes per frame (typical).

MLS and Matrix (OWN) protocol.

Chapters

How can I get end to end encryption in a group call?
So the API used for that, and that is by now implemented in all browsers finally, after a couple of years, is called insertable streams or Media streams Encoded Transform. And what it does, it gives you access to the encoded media packets or the encoded media frames, and then you can run your end to end encryption scheme on top of it.
So just before the packet goes over the network, after WebRTC. No, before WebRTC places it in SRTP. So before the encryption of WebRTC itself, we have access in our application to the packets that need to go out, which are the actual. Let's call them the frames.
Yes. They're framed, so it comes before the packetizer. The packetizer will split them up further.
Okay, so what we get is a full frame already encoded before it goes into rtp.
Yes, yes. And that means the overhead of the encryption, which is typically like 32 bytes or so, is added to the frame, which for video is more efficient.
So what ends up being is that if I want to use end to end encryption, I use insertable streams. I grab that callback encrypt any way I like with whatever key I want and then send it over the network.
Yes. In particular in group calls, how to derive the key for each participant is pretty tricky. There are a couple of things like MLS or Matrix OWN protocol which tackles that problem. Don't reinvent your own crypto.
Yeah. So from our point of view of WebRTC, let's say we leave it to the application to deal with and there are solutions there that you should likely just go and adopt.
Correct.
Okay, thank you.
Thank you.