reenable sntrup761

[seschu]

I have noticed that sntrup761 is still part of libqs and that an old generate.yml file also inlcuded sntrup761. Is it still possible to manually add this algo to the oqs provider lib?

[seschu]

What I did sofar is I added this

diff --git a/oqs-template/generate.yml b/oqs-template/generate.yml
index 099d36e..99940f1 100644
--- a/oqs-template/generate.yml
+++ b/oqs-template/generate.yml
@@ -1,6 +1,20 @@
# This is the master document for ID interoperability for KEM IDs, p-hybrid KEM IDs, SIG (O)IDs
# Next free plain KEM ID: 0x0247, p-hybrid: 0x2F47, X-hybrid: 0x2FB2
kems:
+ -
+ family: 'NTRU-Prime'
+ name_group: 'sntrup761'
+ nid: '0x0233'
+ nid_hybrid: '0x2F44'
+ extra_nids:
+ current:
+ - hybrid_group: "x25519"
+ nid: '0x2FB3'
+ old:
+ - implementation_version: Round 3 p384 hybrid
+ nist-round: 3
+ nid: '0x2F33'
+ oqs_alg: 'OQS_KEM_alg_ntruprime_sntrup761'
-
family: 'FrodoKEM'
name_group: 'frodo640aes'

And then I executed python3 generate.py after setting LIBOQS_SRC_DIR.
Then everything seems to work sofar.

Is the assumption right this can then also be used in a TLS like session?

[baentsch]

Yes (to each of the separate questions above).

[dstebila]

Note that it would not be our intention to permanently re-enable the NTRU Prime suite. We only keep strnup761 in liboqs because this specific algorithm is used in OpenSSH. But since there are (to my knowledge) no standards bodies pursuing standardization of NTRU Prime, we would eventaully deprecate strnup761 in liboqs and remove it.

[loganaden]

If I remember correctly, this is due to OpenSSH stating publicly that sntruprime would be supported as a long term KEX. @baentsch @dstebila With the current patent concerns that were mentioned in the IETF, it might be safer to use have sntrup761 enabled for implementors that might have some concerns ?

[loganaden]

@dstebila There is this draft: https://datatracker.ietf.org/doc/draft-josefsson-ntruprime-ssh/

[baentsch]

I'm a bit confused as to why we have this openssh-related discussion in oqs-provider (only serving openssl): Can you please refresh my memory on this, @loganaden ?

[loganaden]

@baentsch agreed. I went off-topic there. Btw, would a PR to enable ntruprime for openssl be considered ? we want to offer both ML-KEM and sntruprime on a server.

[baentsch]

Btw, would a PR to enable ntruprime for openssl be considered ?

I will consider anything coming with a good reason and an assertion to support it (see for example #351) (and passing CI of course :-)