WG - Attributes In Motion - Comparison of Protocol-based Solutions by Decomposition into Atomic Functions
As the NSTIC work program evolves, we can expect to see multiple proposed solutions put forward that promise to deliver a new generation of online services. To the extent that the various solutions are competing for adoption, it will be necessary to evaluate them against each other for relative costs and benefits. Such comparisons will be quite challenging because proposed solutions will be built on disparate and seemingly incommensurable models (architectures, protocol stacks).
What follows is intended as a first step toward an analytical framework that would allow us to meaningfully compare and contrast widely different solutions to given usage scenarios in the general space of web security. To take an example a SAML-based solution to a given problem might initially appear quite orthogonal to UMA-based solution to the same problem. Yet for the reasons cited above it will be important to be able to evaluate and compare one against the other.
A prerequisite step will be to define a spanning set of atomic functions (technology and protocol-agnostic to the degree possible) that can be shown to be composable in different ways that correspond to familiar protocol-based solution families.
The following is offered as an introductory example. Imagine that a university offers students a service they can use to manage their white-pages entry in the online campus directory. The service allows the student to specify which elements of his/her white pages information should be viewable by anyone and which should be viewable only by faculty, staff and students at institutions within a specified set.
Table I: Atomic functionality required to implement a white pages editing and delivery tool with their composition under two different models:
| Step | Name | Relevant actors or components in SAML model | Relevant actors or components in UMA model |
|---|---|---|---|
| 1 | Request to edit one's own protected White Page (WP) information | Person A as end user --> WP Editing App behind SAML SP | Person A as end user --> WP Client App on Resource Server (RS) |
| 2 | Challenge for Identity | AuthN Service fronting SAML IdP --> Person A as end user | Authorization Server (AS) protecting RS --> Person A as end user |
| 3 | Claim Identity | Person A as end user --> AuthN Service fronting SAML IdP | Person A as Resource Owner --> Authorization Server (AS) protecting RS |
| 4 | Verify Claimed Identity | Authentication Service fronting SAML IdP --> Person A as end user | AS protecting RS --> Person A as Resource Owner (RO) |
| 5 | Grant Authorization to edit WP Information | WP Editing App behind SAML SP --> Person A as end user | AS protecting RS --> Person A as RO |
| 6 | Edit WP Information | Person A as end user --> WP Editing App behind SAML SP | Person A as RO --> WP Client App on RS |
| 7 | Set Access Policy for WP Information | (Done on behalf of Person A by IdP admin per attribute release policy) | Person A as RO --> AS |
| 8 | Persist Access Policy for WP Information | SAML Attribute Release Config Files | AS |
| 9 | Make WP Information Available Online | WP App | Resource Server |
| 10 | Discover White Pages for given user | Person B as end user | Service Registration; Person B as Requesting Party |
| 11 | Search/Find Person WP Information | Person B as end user | Person B as Requesting Party |
| 12 | Request Authorization for WP Information Access | Person B as end user | Person B as Requesting Party |
| 13 |
| ||
| 14 | Grant Authorization for WP Information Access per Policy | WP App behind SAML SP | Authorization Server |
| 15 | Show WP Information | WP App | Resource Server or a Client of Resource Server |
This simple example already highlights some differences between a SAML-based solution and an UMA-based solution. Note that functions performed by the WP App in the SAML model are carried out by more than one component in the UMA model. This helps explain the need in the UMA model for a protocol for cooperatively provided services–The Resource Server and Authorization Server need to collaborate to accomplish the usage scenario. Conversely the comparison highlights that some elements of the usage scenario are out of scope with respect to the SAML model. In other words, a full solution would have to be "SAML plus".