pkcs8: Question around decoding `PrivateKeyInfo`

[Daaiid]

Is it intended behavior that after calling:

let key_info = PrivateKeyInfo::try_from(key_der_bytes).unwrap();

that the private_key field is still DER encoded?

I had to debug for a few hours until I realized that I didn't get the raw key from the field but that it was still prefixed with the Tag and Length bytes.

I would gladly open a PR to document this behavior but I wanted to first make sure what I'm seeing is as intended.

For anyone finding this in the future: Calling

let private_key: &[u8] = OctetStringRef::from_der(key_info.private_key).unwrap().into();

gave me the actual private key.

Thanks in advance!

[tarcieri]

The encoding of the private_key field is specific to the particular algorithm.

We offer many crates with algorithm-specific decoders for that field. Which algorithm are you interested in?

If it’s e.g. Ed25519, check out https://docs.rs/ed25519/latest/ed25519/pkcs8/index.html

[Daaiid]

Thanks for the quick response. I see now why I got stuck in this case.

I'm implementing experimental functionality around the new MLDSA PQC signature algorithms and since TLS integration isn't standardized yet I have to write a lot of stuff myself.

The key I was extracting came from OpenSSL together with the experimental oqsprovider.

The information I was lacking comes from RFC 5958, Section 2:

PrivateKey ::= OCTET STRING
                        -- Content varies based on type of key.  The
                        -- algorithm identifier dictates the format of
                        -- the key.

In that case I would suggest leaving some kind of inline breadcrumb to the fact that this field has different encodings. I saw that the relevant RFC document section is part of the struct docs but the doc string on the field is just Private key data. without any mention of different encodings.

Do you agree with my suggestion? If so I would gladly contribute a PR.

[tarcieri]

Sure, it’d be great if you can improve the docs.

Also check out RustCrypto/signatures#891

[tarcieri]

That said, I'm going to close this out