Skip to Content.
Sympa Menu

edugain-discuss - Re: [eduGAIN-discuss] Post-Quantum readiness and crypto-agility for federation aggregates

edugain-discuss AT lists.geant.org

Subject: An open discussion list for topics related to the eduGAIN interfederation service.

List archive

Re: [eduGAIN-discuss] Post-Quantum readiness and crypto-agility for federation aggregates


Chronological Thread 
  • From: Stefan Winter <stefan.winter AT restena.lu>
  • To: Muhammad Farhan SJAUGI <farhan AT sifulan.my>, "edugain-discuss AT lists.geant.org" <edugain-discuss AT lists.geant.org>
  • Subject: Re: [eduGAIN-discuss] Post-Quantum readiness and crypto-agility for federation aggregates
  • Date: Tue, 3 Nov 2026 14:50:44 +0100
  • Authentication-results: mail.geant.org; dkim=pass (2048-bit key; unprotected) header.d=restena.lu header.i= AT restena.lu header.a=rsa-sha256 header.s=20240131 header.b=jlTchmzi; dkim-atps=neutral

Hello,


that is a great question :-)


I can say which bits and pieces to support it, and some where I found that they don't. All that together is still a small subset of "implementations out there".


* openSSL 3.5.x+ supports the generation of keypairs for RSA/ECDSA/ML-DSA/SLH-DSA with no issues

* xmlsec1 1.3.14+ supports making XMLDSIG signatures with such keys (you need to compile with two extra configure switches to get the new algorithms)

* pyFF 3.0.0 supports validating XMLDSIG signatures with such keys (but it cannot yet produce new signed aggregates, GitHub issue here: https://github.com/IdentityPython/pyFF/issues/349 )


* simpleSAMLphp's metarefresh currently supports validating RSA exclusively (GitHub issue here: https://github.com/simplesamlphp/simplesamlphp-module-metarefresh/issues/80 )


These are the things I tested either to positive or negative result. Everything else is yet unknown to me.


One needs to be aware that cryptographic operations happen on many levels. All of the above relates to metadata handling. There is also the actual SAML message signature and encryption (where the spec may need extension or at least creativity to support new algorithms). And there is the transport-layer security you exchange SAML messages over (where no standard for proper hybrid PQC certificates exists yet). 


Also, to actually complete a transaction in SAML, both the IdP and the SP need to understand the chosen algorithm for things to work. To my knowledge, eduGAIN does not currently include information in metadata regarding which entity supports which algorithms, so with growing algorithm variety, it becomes a guessing game whether a transaction actually succeeds.


Greetings,


Stefan Winter


On 11/3/26 14:40, Muhammad Farhan SJAUGI wrote:
Hi Stefan,
Thank you for the information. I am wondering, how is the support from the IdP/SP software? Does shibboleth IdP and SP support PQC and ECDSA algorithms?
Regards
From: edugain-discuss-request AT lists.geant.org <edugain-discuss-request AT lists.geant.org> on behalf of Stefan Winter <edugain-discuss AT lists.geant.org>
Sent: Tuesday, 3 November, 2026 19:00
To: edugain-discuss AT lists.geant.org <edugain-discuss AT lists.geant.org>
Subject: [eduGAIN-discuss] Post-Quantum readiness and crypto-agility for federation aggregates
 
Hello!


A recent discussion on federation metadata signature requirements
(require larger RSA keys soon) got me thinking about whether RSA still
has enough of a future ahead that it is worth "just" increasing key
sizes for federations that are below the cut.

Current recommendations regarding post-quantum readiness rather suggest
to move away from RSA entirely.

Unfortunately, it is less clear where to move *to* instead. ECDSA is on
the same way out that RSA is. PQC algorithms exist but they are
relatively new and it is not unthinkable that one or some will not stand
the test of time.


In these uncertain times, eduID Luxembourg is taking a step towards PQC,
while adding "crypto-agility" where there was none before.


 From now on, we are publishing our federation aggregate with four
distinct signature algorithms: RSA/4096 bit, ECDSA/NIST P-521 curve,
ML-DSA/87 and SLH-DSA/256s. You can find our aggregates here:
https://restena.lu/en/document/354-implement-eduid-metadata


This is maybe not crypto-agile in a classic reading - every signature
stands on its own, with no relation to the others. But:

* as new algorithms become customary, we can simply enlarge the list;

* if an algorithm falls from cryptographers' grace, it can simply be
removed;

* all variants have identical content except for the <ds:Signature>
element. In case any subset of algorithms is /actually/ compromised and
used to sign bogus federation metadata, this becomes detectable by
comparing all four aggregate files and observing that there is a
difference in the payload.


To my knowledge, eduGAIN OT is already working on the ability to
validate all signature variants; if they were to add a four-way
comparison of all the file contents, we could consider this a full
PQC-ready crypto-agile setup.


Happy to hear your thoughts on this.


Greetings,


Stefan Winter




Archive powered by MHonArc 2.6.19+.

Top of Page