[OAUTH-WG] OSCORE

2018-02-07 Thread Hannes Tschofenig
Hi guys, You may be interested to hear that a group of people working on Internet of Things security believe they have found a solution to deal with the challenges we had in protecting HTTP requests/responses. Here is the draft: https://tools.ietf.org/html/draft-ietf-core-object-security-07 (T

[OAUTH-WG] OSCORE

2018-02-07 Thread Göran Selander
Hi Hannes, and all Thanks for the announcement. To be a little bit more precise, the statement is that a CoAP-mappable HTTP message can be mapped to CoAP (using RFC 8075), protected with OSCORE (as specified in the referenced draft) and transported with HTTP (as exemplified in the referenced dra

Re: [OAUTH-WG] OSCORE

2018-02-07 Thread Hannes Tschofenig
Hi Göran, Maybe you can then answer the question whether this is useful / applicable to a HTTP. Asked differently, under what conditions does the OSCORE not work for HTTP. This would help the folks in the group, including me, to determine whether this actually something we should be looking int

Re: [OAUTH-WG] OSCORE

2018-02-07 Thread Göran Selander
Hi Hannes, Including Dave who may want to provide some background to the use case. As I said, this was a proposed construction and was straightforward to include in the draft. I’m not the right person to answer whether this is useful for OAuth, but I’m interested in the answer. Göran On 2018-0

Re: [OAUTH-WG] OSCORE

2018-02-07 Thread Dave Thaler
As Göran said, yes the original rationale was end-to-end communication through proxies where each leg might be CoAP or might be HTTP, the most common case being a single COAP-to-HTTP or HTTP-to-COAP proxy. For the subset of HTTP that is mappable to CoAP (i.e., simple RESTful calls), I'm not awa

Re: [OAUTH-WG] OSCORE

2018-02-07 Thread Hannes Tschofenig
Is there any implementation / prototyping experience with this work, Dave? Here is what we have been working on in the context of OAuth: With OAuth 1.0: https://tools.ietf.org/html/rfc5849 one of the problems there was the fields we computed the digest over were changed by proxies, and other mid

Re: [OAUTH-WG] OSCORE

2018-02-07 Thread Dave Thaler
Oscore was designed by looking at the specs for HTTP/COAP proxying, so that it should work with any such proxy that's compliant to the spec. I'm not aware if there's implementation experience yet, but the key concept is that fields that have to be preserved end-to-end are tunneled through the pr