TC: Small: Deployment Incentives for Secure Internet Routing

尽管经过了十年的研究，但互联网域间路由系统的安全问题远未解决。在很长一段时间里，似乎存在技术可行性的问题；研究集中在设计越来越轻量级的协议，通过减少计算或通信开销，或考虑较弱的安全保证。现在已经很清楚，部署这些协议的挑战不是技术上的可行性，而是激励措施。互联网是一个复杂的分布式系统，不受任何一个实体的控制，而是有数千名参与者，他们有着不同的经济动机。为了应对这一挑战，该项目开发了：(A)用于测量不同网络安全协议的效用的度量，(B)识别最初应被说服或激励采用新的安全协议的互联网中的自治系统(AS)的目标集合的算法，以便互联网的其余部分具有效仿的动机，(C)当安全路由协议部分部署在互联网中时使用安全路由协议的指南，(D)评估在自动化系统逐步部署该议定书时安全改进情况的模型。该项目中的方法结合了领域知识和对真实网络数据的验证，以及受密码学、博弈论和社交网络文献启发的问题表达。

Despite a decade of research, the problem of securing the Internet's interdomain routing system is far from solved. For a long time, it seemed there was a problem of technical feasibility; research focused on designing more and more lightweight protocols, by reducing computational or communication overheads, or considering weaker security guarantees. It has now become clear that the challenge of deploying these protocols is not one of technical feasibility, but one of incentives. The Internet is a complex, distributed system that is not controlled by any one entity, but instead has thousands of players with different economic incentives. The real challenge is to create incentives for these players to band together and deploy a secure routing protocol that will significantly improve the reliability and security of the Internet.To address this challenge, this project develops: (a) Metrics for measuring the utility of different network security protocols, (b) Algorithms that identify a target set of Autonomous Systems (ASes) in the Internet that should be initially convinced or incentivized to adopt the new security protocols, so that the rest of the Internet has the incentive to follow suit, (c) Guidelines for using secure routing protocols when they are partially deployed in the Internet, (d) Models for evaluating improvements in security as ASes gradually deploy the protocol. The approaches in this project combine domain knowledge and validation on real network data, with problem formulations inspired by emerging ideas from cryptography, game theory, and the literature on social networks.