Unable to create a cert

[hackerfactor]

c2patool 0.6.2 is supposed to accept a user-specified cert via environment variable or in the manifest. I have been unable to get this to work.

The documentation for c2patool 0.6.2 lacks details: https://docs.rs/crate/c2patool/0.6.2 the link for Creating and using an X.509 certificate does not exist.

An older version (0.1.2; https://docs.rs/crate/c2patool/0.1.2) gives instructions using openssl, but those don't seem to work anymore.

Here are the commands I am using:

$ ./openssl version
OpenSSL 3.1.3 19 Sep 2023 (Library: OpenSSL 3.1.3 19 Sep 2023)
$ ./openssl req -new -newkey rsa:4096 \
   -sigopt rsa_padding_mode:pss \
   -days 3650 \
   -extensions v3_ca \
   -addext "keyUsage = digitalSignature" \
   -addext "extendedKeyUsage = emailProtection" \
   -nodes -x509 -sha256 -keyout private3.key -out certs3.pem

My manifest.json contains

"alg": "ps256",
"private_key": "private3.key",
"sign_cert": "certs3.pem",

(Also tried with es256, but that also fails.)

I can't tell if I'm doing something wrong, or if this part of c2patool no longer works correctly.

[hackerfactor]

I found this: https://github.com/contentauth/c2patool/issues/114 As noted by Leszko, the source code is written to explicitly forbid self-signed certs. (https://github.com/contentauth/c2pa-rs/blob/d9b077c8790e172d5bc9f23dc17f13df343160f1/sdk/src/cose_validator.rs#L350)

This contradicts the C2PA specification, which repeatedly mentions the use of self-signed certificates: https://c2pa.org/specifications/specifications/1.2/specs/C2PA_Specification.html#_x_509_certificates E.g.,

• "The presence of a self-signed certificate in the parameter MUST NOT cause the update of the set of trust anchors without some out-of-band confirmation." (permits self-signed)
• "The Authority Key Identifier extension must be present in any certificate that is not self-signed." (permits self-signed)

If you comment out the check/rejection of self-signed certs in the c2pa-rs code, then it correctly accepts self-signed certs. However, nobody else using c2patool will be able to validate it unless they apply the same patch.

[masayadoya]

When using the openssl command to create a certificate, the Basic Constraints of the certificate may be set to "CA:TRUE" if the "-extensions v3_ca" option is used. Isn't this likely to cause errors in c2patool?

If you use "usr_cert" instead of "v3_ca" and create a self-signed certificate with the following command, I think the Basic Constraints of the certificate will be set to "CA:FALSE".

$ ./openssl req -new -newkey rsa:4096
-sigopt rsa_padding_mode:pss
-days 3650
-extensions usr_cert
-addext "keyUsage = digitalSignature"
-addext "extendedKeyUsage = emailProtection"
-nodes -x509 -sha256 -keyout private3.key -out certs3.pem

In my testing, there are no errors in c2patool (0.6.2) when using this certificate, and it seems to be successful.

[hackerfactor]

Oh! That's the magic command. Thank you. (This really needs to be documented somewhere.)

[mauricefisher64]

You must not just supply the signing certificate but also the entire certificate chain up to the root. How to generate that is beyond what is reasonable to document. There are good sources available on the web. If you are just doing testing the c2pa-rs SDK supplies the signing certificates and the appropriate certificate chain for every supported signature type.

[iameli-streams]

Here's the command I found that works for generating a secp256r1 key/cert pair; took me about 45 minutes to figure this out and I had to reference this GitHub issue to get there.

openssl req \
  -new -x509 -nodes \
  -newkey ec:<(openssl ecparam -name prime256v1) \
  -extensions usr_cert \
  -addext "keyUsage = digitalSignature" \
  -addext "extendedKeyUsage = emailProtection" \
  -keyout cert.key -out cert.crt -days 3650

It would be entirely unreasonable to document that command somewhere? Doing so would have saved me 45 minutes. You have a page in your documentation titled "Creating and using an X.509 certificate" that does not document how to create an X.509 certificate. Perhaps a good place to include that command?