Daniel J. Dougherty
Our world teems with networked devices. Their configuration exerts an ever-expanding influence on our daily lives. Yet correctly configuring systems, networks, and access-control policies is notoriously difficult, even for trained professionals. Automated static analysis techniques provide a way to both verify a configuration's correctness and explore its implications. One such approach is scenario-finding: showing concrete scenarios that illustrate potential (mis-)behavior. Scenarios even have a benefit to users without technical expertise, as concrete examples can both trigger and improve users' intuition about their system. This thesis describes a concerted research effort toward improving scenario-finding tools for configuration analysis. We developed Margrave, a scenario-finding tool with special features designed for security policies and configurations. Margrave is not tied to any one specific policy language; rather, it provides an intermediate input language as expressive as first-order logic. This flexibility allows Margrave to reason about many different types of policy. We show Margrave in action on Cisco IOS, a common language for configuring firewalls, demonstrating that scenario-finding with Margrave is useful for debugging and validating real-world configurations. This thesis also presents a theorem showing that, for a restricted subclass of first-order logic, if a sentence is satisfiable then there must exist a satisfying scenario no larger than a computable bound. For such sentences scenario-finding is complete: one can be certain that no scenarios are missed by the analysis, provided that one checks up to the computed bound. We demonstrate that many common configurations fall into this subclass and give algorithmic tests for both sentence membership and counting. We have implemented both in Margrave. Aluminum is a tool that eliminates superfluous information in scenarios and allows users' goals to guide which scenarios are displayed. We quantitatively show that our methods of scenario-reduction and exploration are effective and quite efficient in practice. Our work on Aluminum is making its way into other scenario-finding tools. Finally, we describe FlowLog, a language for network programming that we created with analysis in mind. We show that FlowLog can express many common network programs, yet demonstrate that automated analysis and bug-finding for FlowLog are both feasible as well as complete.
Worcester Polytechnic Institute
All authors have granted to WPI a nonexclusive royalty-free license to distribute copies of the work. Copyright is held by the author or authors, with all rights reserved, unless otherwise noted. If you have any questions, please contact email@example.com.
Nelson, T. (2013). First-Order Models for Configuration Analysis. Retrieved from https://digitalcommons.wpi.edu/etd-dissertations/452
static analysis, model finding, security, firewalls, configuration, scenarios, logic