Add DER encoding/decoding support

[caelunshun]

No description provided.

[caelunshun]

I've implemented a quick version of this at https://github.com/caelunshun/rsa-der. I can look into making a PR for this.

[tarcieri]

You might want to check out @alex's asn1 crate (revamp): alex/rust-asn1#54

[alex]

FYI, I intend to merge and release that soon, hopefully the next week. It doesn't yet have encode support though.

…

On Sun, Sep 1, 2019, 2:43 PM Tony Arcieri ***@***.***> wrote: You might want to check out @alex <https://github.com/alex>'s asn1 crate (revamp): alex/rust-asn1#54 <alex/rust-asn1#54> — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#21?email_source=notifications&email_token=AAAAGBGXRDR52DYONHCIPZTQHQEOVA5CNFSM4IKGWUXKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD5UIKHA#issuecomment-526943516>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAAAGBHXMWKJ6ZFPMVHX3KDQHQEOVANCNFSM4IKGWUXA> .

[newpavlov]

Also I think it will be worth to check out the ring code as well (although note the license), it was proposed to split this code into a different crate in briansmith/ring#586, but unfortunately without much success.

[tarcieri]

Re: that, there's derp, but it's only a parser

[roblabla]

In my code I ended up using pem and yasna to parse private keys generated with openssl: https://github.com/MegatonHammer/linkle/blob/npdm/src/format/npdm.rs#L323

The ASN.1 key format is standardized in RFC 3447, Appendix C. Look for "RSAPrivateKey".

[tarcieri]

@roblabla oh joy, I think you're talking about the "SSLeay format". Does it start with this?

-----BEGIN RSA PRIVATE KEY-----

If so, that's a bespoke PEM encoding of PKCS#1 (i.e. RFC 3447) added to SSLeay, the library which would eventually become OpenSSL.

Personally I'd suggest using PKCS#8 (RFC 5208/5958) instead unless you really, really need to support the SSLeay format for something. These keys start with:

-----BEGIN PRIVATE KEY-----

or

-----BEGIN ENCRYPTED PRIVATE KEY-----

PKCS#8 has the advantage of being reusable for things like ECDSA (and even Ed25519).

[roblabla]

Well I mostly parsed what openssl gave me as an output - which is PKCS#1. I suppose we can get openssl to give us a PKCS#8? (Probably using openssl genpkey instead of openssl genrsa). I’ll look into it.

[tarcieri]

If you're using the OpenSSL CLI, the PKCS#8 functionality is under openssl pkcs8

[nickray]

@tarcieri would you know what's the current state of no_std DER? I need to encode some FIDO2 stuff (signatures, maybe certificates), would lightly prefer to not do things too manually, but also not pull in some huge ASN.1 dependency. The things I've surfaced are all std.

cf. heartsucker/derp#6

[tarcieri]

For signatures, you could look at potentially translating the relevant BearSSL C code. That’s what I did in the ecdsa crate.

For certificates, perhaps consider @Geal’s cookie-factory. If you’re talking about X.509 certificates, you might be able to add no_std support to @str4d’s x509.rs:

https://github.com/str4d/x509.rs

[nickray]

For signatures, you could look at potentially translating the relevant BearSSL C code. That’s what I did in the ecdsa crate.

Ah perfect, your conversions there are exactly what I need, will add them to nisty. Thanks!

[tarcieri]

@nickray you might also look at:

https://www.bearssl.org/gitweb/?p=BearSSL;a=blob_plain;f=src/rsa/rsa_pkcs1_sig_pad.c
https://www.bearssl.org/gitweb/?p=BearSSL;a=blob_plain;f=src/rsa/rsa_pss_sig_pad.c

It looks like BearSSL just hardcodes the relevant ASN.1 DER. See also the relevant decoding code here:

https://www.bearssl.org/gitweb/?p=BearSSL;a=blob_plain;f=src/rsa/rsa_pkcs1_sig_unpad.c
https://www.bearssl.org/gitweb/?p=BearSSL;a=blob_plain;f=src/rsa/rsa_pss_sig_unpad.c

[nickray]

I ended up with a manual mashup of your stuff and derp stuff, I only need ECDSA DER encoding (as the certificates can be hardcoded): https://github.com/nickray/usbd-ctaphid/blob/main/src/derpy.rs. Not super-efficient, but at least it's readable (at least for my mind). Very much hoping I don't have to dig much deeper into this, as I find myself wondering how many semi-efficient pseudo-consistent serialization formats mankind has to come up with...

Funny how Microsoft has the best docs I know of on how this is supposed to work:

• https://docs.microsoft.com/en-us/windows/win32/seccertenroll/about-encoded-length-and-value-bytes
• https://docs.microsoft.com/en-us/windows/win32/seccertenroll/about-integer

[dignifiedquire]

I think most of this is fixed with #38

Could folks confirm if their favorite format is supported by this, or if they are still missing one?

[kitlith]

@dignifiedquire afaict that PR exposes parsing PEM files (#20) but does not directly expose parsing DER. You could parse DER by turning it into a PEM and then feeding it through, but it would be nice if you could just feed it the bytes directly. nope nvm sorry about that.

[tarcieri]

Seems like we can close this issue then?