Using Elliptic Curves is better than RSA – but not as easy

I’ve spent a few days down some rabbit holes trying to solve a problem with an Elliptic Curve, and when I found the answer, I remember I had hit this problem before.

Using an Elliptic Curve generated with

RACDCERT ID( IZUSVR ) GENCERT + 
  SUBJECTSDN(CN('CONN2') O('IBM') OU('C2PDFLT')) + 
  ALTNAME(IP(10.1.1.2))- 
  WITHLABEL('CONN2.IZUDFLT'), + 
     NISTECC + 
     SIZE(521) + 
  SIGNWITH(CERTAUTH LABEL('DOCZOSCAEC' ))    NOTAFTER(DATE(2026/01/01)) 

I was getting in a Zowe trace

javax.net.ssl.SSLException: No supported CertificateVerify signature algorithm for EC key

This was with TLSv1.3.

In the TLS handshake from Chrome I could see

Supported Groups (5 groups)
    Supported Group: Reserved (GREASE) (0x9a9a)
    Supported Group: Unknown (0x11ec)
    Supported Group: x25519 (0x001d)
    Supported Group: secp256r1 (0x0017)
    Supported Group: secp384r1 (0x0018)

I’m sure you spotted the problem immediately. The NISTECC SIZE(521) Elliptic Curve has a name of secp521r1, and this is not in the list. At the server, the logic checks that it’s certificate is supported by the groups sent from the client, and so reports

No supported CertificateVerify signature algorithm for EC key

I found Why is secp521r1 no longer supported in Chrome, others?. The article says that the keysize (256) is a strong as keysize(384) – so why bother with keysize(384).

I changed my certificate to keysize(256) and it worked.