Home | Notifications | New Note | Local | Federated | Search | Logout

Note Detail


Reply to @sun@shitposter.world
andros@andros@activity.andros.dev (2026-07-27 22:06:26)
@sun Good question: Fair, but two separate things.
Deniability: correct, and intentional; HMTP trades deniability for unforgeable authenticity, you can't have both at once (same tradeoff signed PGP/S-MIME always had).
Attachment privacy: valid gap as written. The {hash, url, size} scheme doesn't say the bytes are encrypted before storage, so as described, anyone with the hash+URL could fetch them.
Fix is the same trick used for the body: seal the attachment client-side with the same X25519 sealed box before upload, so the hash addresses ciphertext, not plaintext (what Signal/Matrix do for media). Worth noting attachments aren't in the working prototype at all yet… this part was design-only. Please be understanding, I wrote it on Sunday afternoon while having a coffee. 😄
---Reply--- Blurry Moon@sun@shitposter.world (2026-07-27 22:14:00) @andros no problem. I look forward to seeing more about it in the future.
Reply