Distributed Multivisor

ABSTRACT:A resilient host-based defense must proactively adapt to new attacks and support mission success before, during, and after the attack. Before the attack, a host-based defense must implement defensive measures that do not incur significant overhead for each host or for the enterprise infrastructure. During the attack, the defense technology must present a small attack surface and fight through the attack by limiting the extent of damage caused by successful attacks on both itself and the host. After the attack, the defense technology must swiftly restore full functionality to the system in a way that reduces vulnerability to a repeat attack. We have designed a framework from the bare metal up to the user application that isolates components, so any compromises are limited, detectable, and recoverable. We propose to develop a Distributed Multivisor that provides hardware-enforced isolation to individual software components on a host while presenting a mainstream OS (e.g., Linux) interface to the user and application developers. The Distributed Multivisor enables fault isolation, fault detection, fault recovery, and attack avoidance. Most importantly, it supports these features while guaranteeing strict performance guarantees, so mission-critical applications continue to execute. BENEFIT:We expect the full-scope Distributed Multivisor to have immediate and tangible benefit for a number of military systems that rely on host-based defenses. In particular, Distributed Multivisor will help defend host machines from attacks and enable them to fight through the effects of the attacks with minimal disruptions. Augmenting enterprise hosts with the Distributed Multivisor will enable users to securely execute mission-critical applications.