varx/tech<p>Been thinking about what the key-trust part of decentralized "social PKI" should look like.</p><p>The individual part seems straightforward: A public key, and a history of key rotations (where each key signs a message revoking itself as latest, and endorsing the next public key as latest instead.)</p><p>But how do you trust someone else's key in the first place? (Including when they lose all data and have to recreate their identity, with a new key.)</p><p>I'm very tempted to say... we can leave that to implementations, with just some strong suggestions in the spec as to how to meet different users' different needs.</p><p>One implementation could just use TOFU and notify the user if something looks wrong. Another could participate in a key-gossip system, where useragents inform each other of identity/key relationships they've seen, allowing multipath resilience against MITM. And another could go full-on PGP key-party if it really wanted to, I guess.</p><p>Does this sound reasonable? Would love to hear feedback.</p><p>:boost_ok: </p><p><a href="https://infosec.exchange/tags/cryptography" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>cryptography</span></a> <a href="https://infosec.exchange/tags/SocialMediaDesign" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>SocialMediaDesign</span></a> <a href="https://infosec.exchange/tags/PKI" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>PKI</span></a></p>