Jericho Forum and the Collaboration Oriented Architecture (COA) position paper

Blogger: Dan Blum

After discussing the concept of collaboration oriented architecture (COA) for some time, Jericho Forum released its COA position paper last month at the RSA and Infosecurity Europe conferences. The paper is now posted at http://www.opengroup.org/jericho/COA_v1.0.pdf.

For those who may be unfamiliar with Jericho Forum, it started as a user forum for discussing the problem of deperimeterization, wherein centralized firewalls become less effective as the mainstay of corporate security due to mobility, partnering, outsourcing, telecommuting and all those good things that happen as organizations become more geographically distributed and virtual.

The COA paper focuses on the need for business processes to operate across and between multiple organizations, potentially over untrusted networks such as the Internet. Users and endpoints must securely interact with services and applications controlled by multiple security domains.

The COA position paper builds on the Jericho Forum commandments, which are published at http://www.opengroup.org/jericho/commandments_v1.2.pdf. When reading the commandments, by the way, I find it helps to ignore the explanatory paragraphs, and just focus on the 11 statements of principle. This gets me away from nitpicking the explanations to death and into a state where I just accept them as a very good list of principles for operating securely over open networks.

The COA position paper spends much of its space describing the need for secure, open collaboration as well as principles, processes, standards and frameworks through which the collaboration might be achieved. Most of this doesn’t convey much new information to persons who already grasp the notion of deperimeterization and understand that security is about people, process and technology. But there were some really interesting bits in the section Recommended Solution/Response:

"The COA framework generalizes conventional architectures as follows. It provides:

• increased emphasis on the requirements listed under ‘principles’ below. These are traditionally only seen as external or ‘boundary’ interface concerns in enterprise architectures.
• a user repository (keyed on people identifiers) is generalized into a contract repository (keyed on relationship, or obligation identifiers). A contract repository records agreements, and the obligations and capabilities that ensue from them.
• an accounting log (keyed on system events) is generalized into a reputation repository (keyed on business events). A reputation repository records user actions and compares them to applicable contracts, and, depending on whether or not the actions are in accordance with the contract, upgrades or downgrades a reputation.

The architecture formed by combining SOA (Service Oriented Architecture) with available security protocols (SAML or other XML) is insufficient to support COA. The following elements are also valuable:  [Here, I shorten and paraphrase the list of bullet points]

• attribute brokers
• access brokers
• contract brokers
• policy language (like XACML 3.0)
• performance manager (builds audit logs and reputation systems)”

I wish that the COA position paper had spent more space discussing some of its recommended solutions. The notion of a reputation system (not just a repository) is something we’re hearing more and more about. There is also a growing awareness of the importance of intermediaries, or brokers, that can fairly represent the interests of multiple parties. Perhaps we’ll see this covered in some future Jericho Forum work.

PS: The last bit of COA, in the conclusion, was quite entertaining: “A fundamental shift in thinking is required to implement a COA, moving from the thinking of a hedgehog, an animal that rolls into a tight ball at any sign of threat, to that of a Strawberry Plant, which puts all its key genetic material securely on its outside, as well as sending out suckers to extend the plant’s domain