You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hi AlexLaroche, historically Microsoft has disallowed the use of EdDSA in our products because it was not a FIPS-approved algorithm. As of 2023, EdDSA is now FIPS approved, but we're now in the process of implementing and adopting PQC algorithms such as ML-DSA, so it remains a low priority for us. As such, I do not expect that we'll implement EdDSA any time soon, if ever.
I understand that priority is given to the implementation of post-quantum algorithms.
At the same time, traditional cryptography is here to stay for a while since there is no consensus on how to make the transition between different countries.
In this context, hybridization is at my opinion the best option considering the lack of history of post quantum algorithms in general (except LMS and XMSS).
The case of SIKE is a good example of the lack of perspective on the new post quantum algorithms. (It was a round 4 candidate.)
Are there any plans to implement at least X25519 which is considered in several TLS 1.3 hybridization ciphersuites and used in X-Wing?
Adding a concrete use case from the SymCrypt backend in russh (Rust SSH client/server):
ssh-ed25519 is the default SSH key type: ssh-keygen has defaulted to ssh-ed25519 since OpenSSH 9.5, and ssh-ed25519 (plus ssh-ed25519-cert-v01@openssh.com) is the most widely deployed algorithm for SSH user keys, CAs, and host keys. Even when SSH sessions negotiate post-quantum key exchange (mlkem768x25519-sha256, which we run on SymCrypt's MlKem768 + Curve25519), host and user authentication still rely on ssh-ed25519 / ECDSA / RSA signatures.
Single FIPS module: Because Ed25519 is FIPS-approved under FIPS 186-5, workloads using SymCrypt on Azure Linux and Windows need Ed25519 inside libsymcrypt to stay within a single cryptographic module. Today, supporting ssh-ed25519 requires linking a second C crypto library (aws-lc-sys) solely for Ed25519 (cataggar/russh#62).
Existing building blocks: SymCrypt already implements twisted Edwards curve arithmetic (lib/ec_twisted_edwards.c, SYMCRYPT_ECURVE_TYPE_TWISTED_EDWARDS), SymCryptSha512, and Curve25519 subgroup-order arithmetic, so exposing RFC 8032 / FIPS 186-5 Ed25519 keygen, sign, and verify would close the last algorithm gap for SSH on SymCrypt.
Activity
mlindgren commentedon Jan 9, 2025
Hi AlexLaroche, historically Microsoft has disallowed the use of EdDSA in our products because it was not a FIPS-approved algorithm. As of 2023, EdDSA is now FIPS approved, but we're now in the process of implementing and adopting PQC algorithms such as ML-DSA, so it remains a low priority for us. As such, I do not expect that we'll implement EdDSA any time soon, if ever.
AlexLaroche commentedon Jan 9, 2025
I understand that priority is given to the implementation of post-quantum algorithms.
At the same time, traditional cryptography is here to stay for a while since there is no consensus on how to make the transition between different countries.
Hybrids required: BSI (Germany), ANSSI (France)
Hybrids allowed: ENISA (EU), ETSI
Hybrids discouraged: NSA (US), NCSC (UK), CSE (Canada)
In this context, hybridization is at my opinion the best option considering the lack of history of post quantum algorithms in general (except LMS and XMSS).
The case of SIKE is a good example of the lack of perspective on the new post quantum algorithms. (It was a round 4 candidate.)
Are there any plans to implement at least X25519 which is considered in several TLS 1.3 hybridization ciphersuites and used in X-Wing?
samuel-lee-msft commentedon Jan 14, 2025
X25519 is already supported with
SymCryptEcDhSecretAgreement(usingSymCryptEcurveParamsCurve25519)feat(symcrypt): opt-in Ed25519 from aws-lc-rs (ed25519-aws-lc) (#63)
cataggar commentedon Oct 15, 2026
Adding a concrete use case from the SymCrypt backend in
russh(Rust SSH client/server):ssh-ed25519is the default SSH key type:ssh-keygenhas defaulted tossh-ed25519since OpenSSH 9.5, andssh-ed25519(plusssh-ed25519-cert-v01@openssh.com) is the most widely deployed algorithm for SSH user keys, CAs, and host keys. Even when SSH sessions negotiate post-quantum key exchange (mlkem768x25519-sha256, which we run on SymCrypt'sMlKem768+Curve25519), host and user authentication still rely onssh-ed25519/ ECDSA / RSA signatures.libsymcryptto stay within a single cryptographic module. Today, supportingssh-ed25519requires linking a second C crypto library (aws-lc-sys) solely for Ed25519 (cataggar/russh#62).lib/ec_twisted_edwards.c,SYMCRYPT_ECURVE_TYPE_TWISTED_EDWARDS),SymCryptSha512, and Curve25519 subgroup-order arithmetic, so exposing RFC 8032 / FIPS 186-5 Ed25519 keygen, sign, and verify would close the last algorithm gap for SSH on SymCrypt.cataggar commentedon Oct 16, 2026
Nevermind. After further review, ed25519 is not allowed in my target environment.