No abstract

Abstract. As the number of network-based attacks increase, and system administrators become overwhelmed with Intrusion Detection System (IDS) alerts, systems that respond to these attacks are rapidly becoming a key area of research. Current response solutions are either localized to individual hosts, or focus on a refined set of possible attacks or resources, which emulate many features of low level IDS sensors.In this paper, we describe a modular network-based response framework that can incorporate existing response solutions and IDS sensors. This framework combines these components by uniting models that represent: events that affect the state of the system, the detection capabilities of sensors, the response capabilities of response agents, and the conditions that represent system policy. Linking these models provides a foundation for generating responses that can best satisfy policy, given the perceived system state and the capabilities of sensors and response agents.

Abstract:The goal of any intrusion detection, anti-virus, firewall or other security mechanism is not simply to stop attacks, but to protect a computing resource so that the resource can continue to perform its function. A computing resource, however, is only a component of a larger system and mission. Sometimes, the efforts made to stop an attack on a resource may be as bad as the attack itself in terms of affecting the overall ability of the system to complete its mission. What is needed is a method of choosing responses to attacks on components that still allows the system to achieve its goals. We present a model of computing resources and of how the loss or degradation of resources impacts the ability of a system to complete its mission. A human or robot analyst can use the model to assess the security status of a monitored system and to allocate resources in an optimal way.