Dear authors,

thanks for the draft. I have a couple of comments:

- I agree with your answer in previous mail to the list that "nonce/IV MUST
be random" is not needed. Indeed, a recommendation is more appropriate, as
for a nonce (and AES-GCM-SIV is a nonce-based scheme), it is only important
that it is a *n*umber used *once*.
- I think renaming the field to "Nonce" may be confusing. The field in
IKEv2/ESP context is called IV (for whatever reason) also for other AEAD
modes. Then, one constructs a nonce in some way (N := salt || IV, or in
what you're suggesting N := IV) for the underlying AEAD algorithm and
cryptographic library API.
- In Section 6, you seem to have reversed the order in a sentence. You
write "... its encryption runs at about two thirds the speed of AES-GCM
encryption.". If we're talking about two thirds, then it can only be that
AES-GCM encryption has 2/3 the speed of AES-GCM-SIV (since AES-GCM-SIV
encryption is a sequential two-pass process). Looking at my personal
favorite for AES-GCM(-SIV) benchmarks [1,2], the revised wording roughly
holds true for longer messages, if you want to stick to the "two thirds"
wording.

Cheers,
Vukasin Karadzic

[1]
https://github.com/facebookincubator/haberdashery/blob/main/benchmark_data/md/haberdashery_aes256gcm_skylakex.md
[2]
https://github.com/facebookincubator/haberdashery/blob/main/benchmark_data/md/haberdashery_aes256gcmsiv_skylakex.md
_______________________________________________
IPsec mailing list -- [email protected]
To unsubscribe send an email to [email protected]

Reply via email to