Encryption-Decryption interface for CRYSTALS-Kyber

[iteash]

Purpose: Find the Encryption/Decryption interface for the algorithm CRYSTALS-Kyber

Learning: Only Key Encapsulation/Decapsulation interface is provided by liboqs. Internally there is an indcca enc/dec function that is used but was not able to utilize that functionality properly.

Question: Is there an enc/dec interface provided by Liboqs. If not, does anyone know the rationale behind not providing this interface. In my learning it should be possible to do enc/dec operations using CRYSTALS-Kyber algo. Is my understanding incorrect? Is the enc/dec operation in pipeline and not yet taken up or there is some other fundamental reason for not having this interface.

[baentsch]

If not, does anyone know the rationale behind not providing this interface.

One goal for liboqs used to be to make available interfaces valid for/implementable by all PQ algorithms supported by liboqs.

Personally, I'd think it may be time to revisit this original goal as the set of algorithms supported by liboqs is shrinking and the "downstream" need for an encryption interface clearly exists. What do you think @dstebila @christianpaquin ?

Would you be willing to contribute an API proposal @iteash ? Ideally combined with a use (test!) case...

[dstebila]

Well there are several design choices to be made when converting a KEM into a PKE. I'm not aware of anyone having written out a spec for doing so for Kyber. Probably the right way to do this is plugging Kyber into HPKE. I'm not sure if the right place to do this is in liboqs, or at a higher level. For example, does OpenSSL implement HPKE and can an OpenSSL provider register new HPKE algorithms?

[greengalaxy2016]

I've been thinking of plugging KEM algorithm in to JDK crypto package. that was why in another discussion thread, I asked if supporting OpenJDK is on the plan or not.

[baentsch]

For example, does OpenSSL implement HPKE and can an OpenSSL provider register new HPKE algorithms?

Not that I knew (see supported provider operations). But we could surely work to change that if there were applications for this... Suggestions welcome, @iteash @igorbarshteyn

[iteash]

Hi Michael, Douglas, First and foremost, sorry for the late reply. Our team would love to build a proposal to plug Kyber into a PKE/HPKE scheme if this has not been started yet. Dr. Amit Kumar Chauhan addressed in the mail will be taking it forward from our side. Thanks and regards, Iteash

…

On Tue, Jun 13, 2023 at 1:55 PM Michael Baentsch ***@***.***> wrote: For example, does OpenSSL implement HPKE and can an OpenSSL provider register new HPKE algorithms? Not that I knew (see supported provider operations <https://www.openssl.org/docs/manmaster/man7/provider.html#Operations>). But we could surely work to change that if there were applications for this... Suggestions welcome, @iteash <https://github.com/iteash> @igorbarshteyn <https://github.com/igorbarshteyn> — Reply to this email directly, view it on GitHub <#1464 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ANKCHEFS2GUVWAWWKKILA43XLAPZPANCNFSM6AAAAAAYDG5LJU> . You are receiving this because you were mentioned.Message ID: ***@***.*** com>

[emendir]

I would hugely welcome an interface for basic asymmetric encryption in liboqs.
Not higher-level constructs such as PKE or HPKE, but pure and simple asymmetric encryption.

It's great to have classes for the more complex hybrid encryption ready-to-go, but I find it quite a let-down that the basic, fundamental underlying asymmetric encryption functionality implementing the quantum-resistant algorithms, which is after all, what this library is about, is unavailable in the APIs.

I know most applications use hybrid encryption instead of pure asymmetric encryption, but there are plenty of use-cases for pure asymmetric encryption.

Furthermore, for educational purposes, I think it is vital that users are given access to the fundamental mechanisms underlying higher-level complex constructs such as KEMs.

Adding access to the basic quatum-resistant asymmetric encryption and decryption functions would also make the API more wholesome and complete.
I was quite surprised and disappointed when looking at this cryptography library to see that it has no encrypt and decrypt functions in its API!