I wanted to use encrypted local PASETO tokens for the Request Object
in the OIDC, but it turns out that I need to store the key somewhere to decrypt this request object, and it must be available unencrypted/unhashed as it will need the Request Object
to decrypt. So I will have to store it as plain text in a database? Pretty dangerous. So I started to wonder how JWE works, but the documentation from https://datatracker.ietf.org/doc/html/rfc7516#section-5.1 about JWE encryption is quite confusing for me. Does JWE solve this problem of storing a symmetric key in a database as plain text or does it have other ways?
There are a few different solutions here, which solve different problems:
ENCRYPTED JWTs
These can be used when the app wants to prevent information disclosure. They are issued by the Authorization Server, which uses a public key to encrypt them. There is then a burden on the app to maintain a private key to decrypt them. See the Encrypted ID Tokens for some example usage.
REQUEST OBJECTS
These are often used to protect against man in the browser
tampering. The app only needs to deal with public keys, which it already has access to, so the solution is easier to manage. Newer standards such as PAR
and JARM
are used, as in this summary.
JWT INFORMATION DISCLOSURE
If you want to avoid revealing sensitive data in access token JWTs, then the usual technique is to return only opaque access tokens to internet clients. This is easier to manage than encryption. See the Phantom Token Pattern for how this works.
SUMMARY
I would usually avoid introducing key management into apps. Aim to manage this in the Authorization Server instead.