Policy management and interoperation through negotiation in ubiquitous computing.

摘要

Ubiquitous computing environments consist of autonomous domains that are not administered centrally and have independent goals and policies. These domains, which could be single mobile devices or networks of devices, are capable of communicating through standardized protocols and identifying external services. But true spontaneous interoperation leading to policy-compliant resource and service access agreements across domains is yet to be realized. The large number of possible interaction contexts, resource heterogeneity, and differing security policies of the domains make the use of application level protocols for every scenario impractical and non-scalable. Also, every domain cannot expect to identify or have a pre-arranged trust relationship with every other domain.;We have designed and implemented a generic negotiation protocol based on illocutionary speech acts that enables domains to reach resource and service access agreements. This protocol is guided by the local private policies of each domain, which specify system invariants, goals, resource usage, and security constraints, in a declarative logical language. Negotiation was achieved within a broader policy management and mediation framework, which was designed and implemented as part of the Panoply ubiquitous computing middleware. This framework also provides other services, including dynamic context-sensitive access control through message filtering and event-driven system responses.;In this dissertation, I will describe mobile and ubicomp applications that benefit from policy management and negotiation, and show that the protocol performance is adequate for practical scenarios. I will show how the negotiation protocol was modeled as a distributed policy resolution, where neither negotiator is privy to the other's policies, and analyze its theoretical correctness properties. I will describe how random test cases were generated for the purpose of comparing negotiation performance to centralized policy resolution that is optimal in the number of negotiation steps. The results indicate the feasibility of using the negotiation protocol to generate agreements in ubicomp scenarios.