[secdir] SECDIR review of draft-ietf-oauth-jwsreq-09.txt

I generated this review of this document as part of the security directorate's ongoing effort to review all IETF documents being processed by the IESG.  These comments were written with the intent of improving security requirements and considerations in IETF drafts.  Comments not addressed in last call may be included in AD reviews during the IESG review.  Document editors and WG chairs should treat these comments just like any other last call comments.

This document proposes a mechanism to enable secure communication of OAuth 2.0 Authorization Requests using a JSON Web Token (JWT). This mechanism represents an improvement over the current way that OAuth Authorization Requests are transmitted, i.e., encoded as an (unprotected) URI.

The document notes that the current Authorization Request mechanism fails to provide integrity, authentic, or confidentiality. JSON is already used for OAuth responses, so using JWT to protect requests seems like an appropriate choice. (XML signatures and encryption were rejected as too complex.)

Section 4 defines the Request Object format and provides examples.

The text here is a bit confusing. It seems to state that only integrity and authenticity are mandated by this specification; confidentiality is an optional feature. However, when discussing the use of encryption that does not provide authentication, the text says that a signature “should” (not SHOULD””) be applied. The text then says that “In this case, it [the token] MUST be signed then encrypted …” This combination of sentences is confusing and OUGHT J to be revised.

Section 6 describes how to validate a received JWT request token. Section 6.1 appears to not mandate use of a signature for an encrypted token, suggesting that authentication and integrity need not be provided if the requestor encrypts the token (and does not employ an authenticated encryption algorithm).

Section 10 describes Security Considerations in addition to the ones already describes in RFC 6119 (OAuth 2.0). The wording of Section 10.1 is odd: “ …it MUST either be JWS signed with then considered appropriate algorithm or encrypted using [RFC7516].” Why is there no cite of 7515 for JWS algorithms here, to parallel the cite of JWE?

Section 10.2 indicates that a client and server might agree, a priori, to use the non-protected parameters transmitted in a request. It does not indicate how this might have been done (hopefully, in a secure fashion).

Section 10.3 finally mandates authentication of the request source, something that was ambiguous in earlier sections of this document. There are some ambiguous statement here, e.g. “Since Request Object URI can be replayed, the lifetime of the Request Object URI MUST be short and preferably one-time use.  The entropy of the Request Object URI MUST be sufficiently large.” The lack of guidance of what constitutes a “short” lifetime or a “sufficiently large” amount of entropy (in a short URI) is worrisome.  In (d) there is a typo: “The same requirements as (b) above applies.” -> “The same requirements as (b) above apply”.

Section 10.4 includes several typos:

“Although this specification does not require them, researchs such as …” -> “Although this specification does not require them, research such as …” This is the beginning of a run-on sentence.

“The endpoints that comes into question …” -> The endpoints that come into question …”

The wording in several places is awkward, e.g., missing articles.

This section ends with the statement “An extension specification should be created.” Presumably the intent here is to suggest that an extension is needed to remedy the vulnerability resulting from the lack of explicit endpoint identifiers. This should be more clearly stated.

Section 11 discusses Privacy Considerations an unusual element of an RFC. (The authors state that ISO/IEC 29100 is freely accessible. That seems to be true only if one follows the URL in the Informative References. A search for this ISO document tends to yield copies available for a non-trivial fee, i.e., ~ $150 USD.) Since there is standards language in this section (SHOULD and MUST) I think 29100 needs to be a Normative (not Informational) reference.

The text here raises some good privacy concerns and suggests some means by which these concerns might be addressed. However, the wording here needs to be significantly improved. There are extraneous articles and missing articles that make the text harder to read. The ambiguous comment about entropy that appeared in 10.3 appears here as well.