Preventing denial and forgery of causal relationships in distributed systems

In a distributed system, it is often important to detect the causal relationships between events, where event e1 is causally before event e2 if e1 happened before e2 and could possibly have affected the occurrence of e2. In this paper we argue that detecting causal relationships among events can be important for security, in the sense that it may be essential to the correct implementation of a security policy that a process be able to determine if two events are causally related, and if so, how. We formulate attacks on causality in terms of causal denial and forgery, formalize possible security goals with respect to causality, and present simple algorithms to attain these goals in some situations.

Duke Scholars

Author Michael Reiter Computer Science