| dc.description.abstract | Firewalls, network devices, and the access control lists that manage traffic are very important components of modern networking from a security and regulatory perspective. When computers were first connected, they were communicating with trusted peers and nefarious intentions were neither recognized nor important. However, as the reach of networks expanded, systems could no longer be certain whether the peer could be trusted or that their intentions were good. Therefore, a couple of decades ago, near the widespread adoption of the Internet, a new network device became a very important part of the landscape, i.e., the firewall with the access control list (ACL) router. These devices became the sentries to an organization's internal network, still allowing some communication; however, in a controlled and audited manner. It was during this time that the widespread expansion of the firewall spawned significant research into the science of deterministically controlling access, as fast as possible. However, the success of the firewall in securing the enterprise led to an ever increasing complexity in the firewall as the networks became more inter-connected. Over time, the complexity has continued to increase, yielding a difficulty in understanding the allowed access of a particular device. As a result of this success, firewalls are one of the most important devices used in network security. They provide the protection between networks that only wish to communicate over an explicit set of channels, expressed through the protocols, traveling over the network. These explicit channels are described and implemented in a firewall using a set of rules, where the firewall implements the will of the organization through these rules, also called a firewall policy. In small test environments and networks, firewall policies may be easy to comprehend and understand; however, in real world organizations these devices and policies must be capable of handling large amounts of traffic traversing hundreds or thousands of rules in a particular policy. Added to that complexity is the tendency of a policy to grow substantially more complex over time; and the result is often unintended mistakes in comprehending the complex policy, possibly leading to security breaches. Therefore, the need for an organization to unerringly and deterministically understand what traffic is allowed through a firewall, while being presented with hundreds or thousands of rules and routes, is imperative. In addition to the local security policy represented in a firewall, the modern firewall and filtering router involve more than simply deciding if a packet should pass through a security policy. Routing decisions through multiple network interfaces involving vendor-specific constructs such as zones, domains, virtual routing tables, and multiple security policies have become the more common type of device found in the industry today. In the past, network devices were separated by functional area (ACL, router, switch, etc.). The more recent trend has been for these capabilities to converge and blend creating a device that goes far beyond the straight-forward access control list. This dissertation investigates the comprehension of traffic flow through these complex devices by focusing on the following research topics: - Expands on how a security policy may be processed by decoupling the original rules from the policy, and instead allow a holistic understanding of the solution space being represented. This means taking a set of constraints on access (i.e., firewall rules), synthesizing them into a model that represents an accept and deny space that can be quickly and accurately analyzed. - Introduces a new set of data structures and algorithms collectively referred to as a Firewall Policy Diagram (FPD). A structure that is capable of modeling Internet Protocol version 4 packet (IPv4) solution space in memory efficient, mathematically set-based entities. Using the FPD we are capable of answering difficult questions such as: what access is allowed by one policy over another, what is the difference in spaces, and how to efficiently parse the data structure that represents the large search space. The search space can be as large as 288; representing the total values available to the source IP address (232), destination IP address (232), destination port (216), and protocol (28). The fields represent the available bits of an IPv4 packet as defined by the Open Systems Interconnection (OSI) model. Notably, only the header fields that are necessary for this research are taken into account and not every available IPv4 header value. - Presents a concise, precise, and descriptive language called Firewall Policy Query Language (FPQL) as a mechanism to explore the space. FPQL is a Backus Normal Form (Backus-Naur Form) (BNF) compatible notation for a query language to do just that sort of exploration. It looks to translate concise representations of what the end user needs to know about the solution space, and extract the information from the underlying data structures. - Finally, this dissertation presents a behavioral model of the capabilities found in firewall type devices and a process for taking vendor-specific nuances to a common implementation. This includes understanding interfaces, routes, rules, translation, and policies; and modeling them in a consistent manner such that the many different vendor implementations may be compared to each other. | |
