Re: [Rift] RIFT fingerprint coverage

[Bruno Rijsman <[e-mail]>]

> On Jul 21, 2019, at 4:22 PM, Antoni Przygienda <[e-mail]> wrote:

> 2. first, you can't extend the protection the way you like unless we have a complicated procedure where you protec with fingerprint 0, then you write fingerprint and on the other side you ahve to 0 out the fingerprint to check and so on. I wrote such code over protocols and it's very fragile. 


I think we should focus on whether it is a security issue or not (which was #3 of your response) rather than whether the implementation is difficult or not.

That said, the implementation is actually quite trivial and not fragile. 

Every signing library that I ever worked with has an update method that allows you to incrementally compute a digest.

Python provides hmac.update() and hashlib.update(). Rust provides hmac.input().

Python example:

import hmac
packet = b'Just imagine for a second that this is a protocol packet.'
signer = hmac.new(key=b'top-secret', digestmod='sha256')
signer.update(packet[5:10])  # Digest over bytes 5-10
signer.update(packet[15:20]) # and also bytes 15-20
signer.update(packet[25:30]) # and also bytes 25-30
print(signer.digest())

You can use this feature to make re-computing the outer fingerprint for TIEs that are stored in the TIE database much cheaper, knowing that you have already computed (and saved the signer for) the inner fingerprint.