I had tried to address this in the security considerations section of the 
draft. I have not submitted the next version yet, but here is the text

~~~
For post-quantum deployments where the responder is upgraded to support ML-KEM 
before any initiator, the initiator could enforce a hard requirement for using 
ML-KEM in the IKE_SA_INIT or the IKE_INTERMEDIATE before encrypting any data in 
any Child SA or fail the negotiation. This is the most straightforward 
protection against downgrades in cases where the responder is upgraded before 
any initiator. Otherwise, [RFC9370] supports Childless IKE SAs which can be 
followed by a Child SA after a FOLLOWUP_KE exchange with ML-KEM. Establishing a 
Childless IKE SA or a Child SA which does not encrypt any data and establishing 
a Child SA or rekeying the existing one with a FOLLOWUP_KE exchange with ML-KEM 
ensures that the initiator or responder can enforce a policy which requires 
ML-KEM for peers expected to support ML-KEM after learning their identity in 
IKE_AUTH. Although this approach would prevent downgrades where an adversary 
can force peers that support ML-KEM to use classical key exchanges, it assumes 
the initiator or responder know the identities of their peers that support 
ML-KEM. It also has the disadvantage that an adversary with a quantum computer 
that can decrypt the initial IKE SA has access to all the information exchanged 
over it, such as identities of the peers, configuration parameters, and all 
negotiated SA information (including traffic selectors), with the exception of 
the cryptographic keys used by the IPsec SAs which are established after the 
ML-KEM key exchange.
~~~

From: Jun Hu (Nokia) <[email protected]>
Sent: Monday, July 21, 2025 10:07 AM
To: Christopher Patton <[email protected]>; Valery Smyslov 
<[email protected]>
Cc: Daniel Van Geest 
<[email protected]>; [email protected]
Subject: [EXTERNAL] [IPsec] Re: Binding properties of 
draft-ietf-ipsecme-ikev2-mlkem-00


CAUTION: This email originated from outside of the organization. Do not click 
links or open attachments unless you can confirm the sender and know the 
content is safe.


Clarifying question: How exactly would it work to disable weak KEs for peers 
that support strong KE? The peer doesn't identify itself until the IKE_AUTH 
exchange, at which point the sequence of KEs has already been negotiated and 
executed. Is it possible to abort due to insufficient KE parameters at this 
point?
[HJ] as described in my previous email, in typical deployments, both peers 
either belong to or controlled by same organization, so at least one peer will 
be able to know remote peer’s capabilities without relying on IKE negotiation, 
and could have corresponding configuration to disable weak KEs toward that 
peer; another approach is like you mentioned, there is IPsec implementation 
could abort negotiation base on local policy config once learned peer’s IKEv2 ID

_______________________________________________
IPsec mailing list -- [email protected]
To unsubscribe send an email to [email protected]

Reply via email to