News for hackers. Stuff that matters.
Re:Rolling Your Own Crypto by danbruc on Tuesday December 20, 2016 @04:54PM (#13224461)
Poly1305 is one of the few primitives out there that are provably secure.

That doesn't sound right. I guess this should read something like »is provably as hard to break as the underlying block cipher« in the same way that that a hash function built using the Merkle-Damgård construction is provably as hard to break as the compression function used. The article easily gives the impression that Poly1305 is provably secure in the same way as one-time pad.

Re:Rolling Your Own Crypto by loup-vaillant on Thursday December 22, 2016 @07:44AM (#13236882)
I have thought about it and re-read the original paper. I stand by my claim.

The truth of the matter is, poly1305, while provably secure, is almost as impractical as a one time pad: it relies on a shared random authentication key. Where does that key is supposed to come from? In practice, you'd derive it from a session key (and so rely on a symmetric cipher such as AES or Chacha), or from a key exchange scheme such as Diffie-Hellman, and you rely on that.

I left my phrasing as it was, because it is (I think) closer to the truth, and not dangerous. I do reckon the AEAD scheme I recommend only has a reduction proof, not a safety proof. I could be more precise, but I don't like clutter.