`verify_certificate_identity` fails when `subjectAltName` includes `otherName`
verify_certificate_identity relies on a naive string-based approach to extract DNS: (and IP:) entries from the subjectAltName extension. However, the SAN is truly an ASN.1 structure, and when the library stringifies it, otherName entries and semicolons (or other delimiters) will appear in the output. This leads to the regex split call misreading the line, causing verify_certificate_identity to fail on certificates that include otherName (common in Active Directory environments, for example).
For instance, a SAN might be stringified like:
"subjectAltName = critical, otherName:[1.3.6.1.4.1.311.20.2.3, [CONTEXT 0][email protected]];DNS:host1.example.com, DNS:example.com, DNS:MYDOMAIN"
(See https://github.com/jruby/jruby-openssl/blob/976a3f5152b36129ad478175473bd63345286450/src/main/java/org/jruby/ext/openssl/X509Extension.java#L518 for where this happens?)
When split(/,\s+/) is applied to this string (see https://github.com/jruby/jruby-openssl/blob/976a3f5152b36129ad478175473bd63345286450/lib/openssl/ssl.rb#L273C9-L273C32), the returned array is
[
"subjectAltName = critical",
"otherName:[1.3.6.1.4.1.311.20.2.3",
"[CONTEXT 0][email protected]];DNS:host1.example.com",
"DNS:example.com", "DNS:MYDOMAIN"
]
The DNS:host1.example.com entry will not be found and extracted by the regex in if /\ADNS:(.*)/ (https://github.com/jruby/jruby-openssl/blob/976a3f5152b36129ad478175473bd63345286450/lib/openssl/ssl.rb#L277C11-L277C26)
@kares In https://github.com/jruby/jruby-openssl/commit/f18c79419c7ad9d12b59deb93f00d34963593a1f it looks like you might have tested the MRI openssl implementation and decided the custom approach was still needed???
There does seem to be an issue using the version from https://github.com/ruby/openssl as-is. Specifically, san.value (https://github.com/ruby/openssl/blob/e5153dbbb4baa568b48bf2caf9af9b79dbfe2f1b/lib/openssl/ssl.rb#L323C54-L323C63) doesn't return a raw string in jruby-openssl compared to ruby-openssl, but a single element Array of OpenSSL::ASN1::OctetString instead??
I've had a go at adding a test and fixing this issue. I'm not confident in my approach though. PR to follow shortly.
and decided the custom approach was still needed???
yes, unfortunately the ASN.1 parsing isn't 100% complete and there are differences in how the ext.value is presented. there were changes (fixes) to ASN.1 since but still brittle IMO.
changes (e.g. "Same here. san.value is changed to san.value.last.value") in your PR is risky to be put in. those changes would require more tests.