Enforcing trust for execution-protection in modern environments

Abstract

The business world is exhibiting a growing dependency on computer systems, their operations and the databases they contain. Unfortunately, it also suffers from an ever growing recurrence of malicious software attacks. Malicious attack vectors are diverse and the computer-security industry is producing an abundance of behavioral-pattern detections to combat the phenomenon. Modern processors contain hardware virtualization capabilities that support implementation of hypervisors for the purpose of managing multiple Virtual-Machines (VMs) on a single computer platform. The facilities provided by hardware virtualization grant the hypervisor control of the hardware platform at an effective privilege level that supersedes the OS. The purpose of this work is to research and develop a methodology based on a thin-hypervisor that exploits the virtues of hardware virtualization for the purpose of protecting a computer system against malicious penetration. To successfully accomplish this, the thin-hypervisor must be guaranteed to be trusted, with respect to its instructions its configuration structures and its true control over the hardware platform. Moreover, it must be able to protect itself indefinitely from subversion. The methodology presented here describes the means to establish a trusted thin-hypervisor and demonstrates how it may be exercised to restrict code execution exclusively to pre-signed, whitelisted, software. This methodology provides resistance to most APT attack vectors, including those based on zero-day vulnerabilities that may slip under behavioral-pattern radars.