OpenID4VCs: Security & Trust Model Document

[Torsten Lodderstedt]

We need a comprehensive analysis and description of the security of the OpenID4VCs protocol family, which also includes the underlying trust model. It is important to conduct the analysis end 2 end for the whole family since there are interdependencies.

Here are just some initial thoughts:

• Authentic claims - signed data by a trusted issuer
• trusted issuer - signed data / identification of issuer / management of trusted issuers
• Prevention of copying/replay of credentials / impersonation - holder binding (cryptographic, biometric, claims-based)
• Security of cryptographic holder binding - attestation of different kind
• Trustworthiness of processing by wallet - 3rd party verification / confirmation
• Design philosophy: issuer does the heavy lift, verifier can be sure all pre-requisites are fulfilled by wallet if there exists a credential issued by a trusted issuer (since it conforms to its policy)
• Confidentially of user claims - protected cloud systems / encryption on rest, encrypted transmission
• Issuance - Do we need nonce and audience in issuance?
• Impersonation by attacker - authentication to wallet

‌

[David W Chadwick]

Specify conceptual models first.

Re: design philosophy. Keep it simple, pin down options to mimimum for viable open interworking.

[Kristina Yasuda]

there is also Issue #1556

for VP spec:

• How does a verifier establish trust in a wallet, esp. its ability to handle key material securely?
• presentation_submission is unsigned metadata, handle with care

‌

[Kristina Yasuda]

also #1516

[Kristina Yasuda]

also #1425

[Joseph Heenan]

As discussed on today’s call, response_mode=post probably opens up some reflection/open proxy style attacks, and Brian commented that the cross-device flow in general likely has more.

[Mark Haine]

Migrated to GitHub

https://github.com/openid/OpenID4VP