Avoiding information leakage in security-policy-aware planning

Abstract

In early computer systems only simple actions would be governed by security policies. However, computers are increasingly handling complex organizational tasks which may have complex preconditions and postconditions. As such, it is useful to be able to plan and schedule actions in advance in order to ensure that desired actions will be able to be carried out without violating the security policy. However, there is a possibility that planning systems could accidentally leak information about future plans which should be kept confidential. In this paper, we investigate how sensitive information could be leaked by a planning system which uses security policies to ensure that planned actions will be able to occur. We formally define information leakage in this context. Then we present two techniques which can be used to mitigate or eliminate this information leakage and prove their security.