Optimal Cost, Collaborative, and Distributed Response to Zero-Day Worms - A Control Theoretic Approach

Collaborative environments present an happy hunting ground for worms due to inherent trust present amongst the peers. We present a novel control-theoretic approach to respond to zero-day worms in a signature independent fashion in a collaborative environment. A federation of collaborating peers share information about anomalies to estimate the presence of a worm and each one of them independently chooses the most cost-optimal response from a given set of responses. This technique is designed to work when the presence of a worm is uncertain. It is unique in that the response is dynamic and self-regulating based on the current environment conditions. Distributed Sequential Hypothesis Testing is used to estimate the extent of worm infection in the environment. Response is formulated as a Dynamic Programming problem with imperfect state information. We present a solution and evaluate it in the presence of an Internet worm attack for various costs of infections and response.