Confidence Domains for Distributed Systems

The paper addresses the problem of trust in large computer networks that connect several independent organizations. While in such networks it is politically difficult to agree upon one single common point of trust and one single global network security policy, few networks exist in which no system trusts any other system. Thus we observe that systems in a network form clusters, based on the sharing of a common point of trust or a common security policy. One of the major assumptions in this paper is that trust cannot be achieved ona simple technical or mechanical level alone. We introduce confidence domains as an approach to describe human belief in the trustworthyness of systems and thus make this knowledge available to the system's security components. The paper describes the concept of confidence domains together with the paradigms used to de ne and estabish them. It gives examples how confidence domains can be exploited as a foundation for security policies. The paper also describes mechanisms needed to enforce confidence domains in an open network and concludes with a detailed description of an implementation for the BirliX Security Architecture