Post #3846256

2026-07-16 00:44 UTC

@soatok@furry.engineer @hazelnoot@enby.life I guess a higher level question: most people use fedi partially or entirely via web frontends. Is any kind of meaningful E2EE possible if you aren't storing secrets and running trusted code client side in a thick client? I think not - the instance either has to store a long lived key or derive it from your password, and even if the crypto itself runs clientside it's in javascript authenticated only by the server's TLS certificate. Meaning that a rogue instance can read or tamper with your traffic. So what's the plan for this use case? Do you make E2EE messages only work with native clients? Allow users to store a key server side and make the UI clearly explain that they are trusting the instance by doing so? I feel like the UI/UX challenges here are 10x harder than the cryptographic ones.

Replies (2)

@azonenberg@ioc.exchange 2026-07-16 00:46

@soatok@furry.engineer @hazelnoot@enby.life or is there some means I'm unaware of by which a web app can store keys clientside and encrypt/decrypt stuff with them in constrained ways like displaying to the user, but prevent malicious JS in the same context from exfiltrating plaintext or keys to the server?

Open ##3846288
@NfNitLoop@mastodon.social 2026-07-16 01:18

@azonenberg@ioc.exchange @soatok@furry.engineer @hazelnoot@enby.life When I was working on #Diskuto (https://github.com/diskuto/) I settled on a UI that performed signing outside of the browser. You can have a little trusted tool that receives the representation of a post and signs it. Paste the signature back in to the browser (which verifies it) and click Post. It's equivalent to signing a message with PGP/GPG before posting it to a forum.

Open ##3847047