Towards Self-Explaining Networks

摘要: In this paper, we argue that networks should be able to explain their operators why they are in a certain state, even if – and particularly have been compromised by an attacker. Such capability would useful forensic investigations, where operator observes unexpected state must decide whether it is benign or indication the system has compromised. Using very pessimistic threat model which malicious adversary can completely compromise arbitrary subset of nodes network, cannot expect get complete correct explanation all possible cases. However, also show that, based on recent advances systems database communities, seems slightly weaker guarantee: for any change directly indirectly affects node, either obtain eventually identify at least one node. We discuss challenges involved building provide property, report initial results from early prototype.