Sharon Goldberg ­­­ Research Statement

As a security researcher, my goal is to develop practical solutions that make computer networks more reliable, predictable, and resilient to attack. To address the grand challenges in security research, I believe that we need to go beyond traditional disciplinary barriers. For instance, cryptography is crucial to deal with hacked devices and malicious parties, but cannot deal with the fact that nodes in computer networks are often owned by entities with competing economic goals. Game theory is an emerging approach that can be used to reason about the way systems behave in the presence of selfish (rational) behavior, but often requires making strong assumptions on the economic motivations of parties in the network. At the end of the day, solutions must be implemented in extremely resource-constrained network devices, so hardware and software engineering concerns are paramount. Finally, ensuring that security research makes a real impact requires an investment in technology transfer activities, including standardization, prototyping or collaboration with industry. Indeed, my research spans all of these areas. When we purchase an item from Amazon.com, traditional cryptography prevents attackers from seeing our credit card numbers or impersonating the Amazon website. But how can we ensure that our request actually arrives at the Amazon.com server, without being dropped or corrupted along the way? This is exactly the challenge we address when we consider securing Internet's interdomain routing system. The interdomain routing system enables communication between autonomous systems (e.g. Princeton's campus network, AT&T's global backbone network, and China Telecom's residential network). Each autonomous system has different economic goals, each controls a sophisticated network of devices, and each is prone to malfunction, misconfiguration, or attack by malicious parties. Despite these complex relationships, the interdomain routing system was originally designed under the assumption that nodes can trust each other. Furthermore, the system is notoriously resistant to change. This is mainly due to the difficulty of coordinating and convincing autonomous systems to upgrade to one of the new security protocols developed over the last decade. To advance efforts to secure the system, my dissertation research addressed the following two questions: Q1. What is the right layer of the system to secure? Q2. What are the right tradeoffs between security and efficiency (e.g. storage and computation overhead)? To address the questions above, I considered separately the effectiveness of securing the following two layers of the interdomain routing system: the routing protocols, used to set up paths through …