Channels cannot be correlated to users unless:
- you leak via IP
- you leak via activity times
Both of which apply to all schemes.
This is only for obscuring who a user has communicated with at all: if an app establishes channels only when DMing, you only leak A has initiated a channel with B, but adding noise reduces this risk.
Ok. I think I get it now. I’ll try and play with it tomorrow.
I’ll be doing some revisions to the spec today for fixing some flaws.
Most importantly, an attacker can request re-Initialization of the channel while only having the channel state, allowing them to pretend to be the public key way after the channel gets cut.
I’ll also be introducing a control command to stop using a channel.
