Flow-Sensitive Semantics for Dynamic Information Flow Policies

Paper in proceeding, 2009

Dynamic information flow policies, such as declassification, are essential for practically useful information flow control systems. However, most systems proposed to date that handle dynamic information flow policies suffer from a common drawback. They build on semantic models of security which are inherently flow insensitive, which means that many simple intuitively secure programs will be considered insecure. In this paper we address this problem in the context of a particular system, flow locks. We provide a new flow sensitive semantics for flow locks based on a knowledge-style definition (following Askarov and Sabelfeld), in which the knowledge gained by an actor observing a program run is constrained according to the flow locks which are open at the time each observation is made. We demonstrate the applicability of the definition in a soundness proof for a simple flow lock type system. We also show how other systems can be encoded using flow locks, as an easy means to provide these systems with flow sensitive semantics.