Has anybody thought about modelling #activitypub with a tool like https://alloytools.org/book.html to find potential exploits? Thinking about the spec it’s missing any algorithms for authorization, but I already found a couple of edge-cases that make a server DoSssable or give an attacker the ability to spoof messages …
@silverpill does the http signature not contain the domain of the requesting server and if yes, can't it be used to compare origins after the signature check?
@silverpill I mean ideally we already have a cache from the corresponding server key to its origin, so we don’t have to do a https resolution on every incoming message