A pragmatic approach toward securing inter-domain routing

Abstract

Internet security poses complex challenges at different levels, where even the basic requirement of availability of Internet connectivity becomes a conundrum sometimes. Recent Internet service disruption events have made the vulnerability of the Internet apparent, and exposed the current limitations of Internet security measures as well. Usually, the main cause of such incidents, even in the presence of the security measures proposed so far, is the unintended or intended exploitation of the loop holes in the protocols that govern the Internet. In this thesis, we focus on the security of two different protocols that were conceived with little or no security mechanisms but play a key role both in the present and the future of the Internet, namely the Border Gateway Protocol (BGP) and the Locator Identifier Separation Protocol (LISP). The BGP protocol, being the de-facto inter-domain routing protocol in the Internet, plays a crucial role in current communications. Due to lack of any intrinsic security mechanism, it is prone to a number of vulnerabilities that can result in partial paralysis of the Internet. In light of this, numerous security strategies were proposed but none of them were pragmatic enough to be widely accepted and only minor security tweaks have found the pathway to be adopted. Even the recent IETF Secure Inter-Domain Routing (SIDR) Working Group (WG) efforts including, the Resource Public Key Infrastructure (RPKI), Route Origin authorizations (ROAs), and BGP Security (BGPSEC) do not address the policy related security issues, such as Route Leaks (RL). Route leaks occur due to violation of the export routing policies among the Autonomous Systems (ASes). Route leaks not only have the potential to cause large scale Internet service disruptions but can result in traffic hijacking as well. In this part of the thesis, we examine the route leak problem and propose pragmatic security methodologies which a) require no changes to the BGP protocol, b) are neither dependent on third party information nor on third party security infrastructure, and c) are self-beneficial regardless of their adoption by other players. Our main contributions in this part of the thesis include a) a theoretical framework, which, under realistic assumptions, enables a domain to autonomously determine if a particular received route advertisement corresponds to a route leak, and b) three incremental detection techniques, namely Cross-Path (CP), Benign Fool Back (BFB), and Reverse Benign Fool Back (R-BFB). Our strength resides in the fact that these detection techniques solely require the analytical usage of in-house control-plane, data-plane and direct neighbor relationships information. We evaluate the performance of the three proposed route leak detection techniques both through real-time experiments as well as using simulations at large scale. Our results show that the proposed detection techniques achieve high success rates for countering route leaks in different scenarios. The motivation behind LISP protocol has shifted over time from solving routing scalability issues in the core Internet to a set of vital use cases for which LISP stands as a technology enabler. The IETF's LISP WG has recently started to work toward securing LISP, but the protocol still lacks end-to-end mechanisms for securing the overall registration process on the mapping system ensuring RLOC authorization and EID authorization. As a result LISP is unprotected against different attacks, such as RLOC spoofing, which can cripple even its basic functionality. For that purpose, in this part of the thesis we address the above mentioned issues and propose practical solutions that counter them. Our solutions take advantage of the low technological inertia of the LISP protocol. The changes proposed for the LISP protocol and the utilization of existing security infrastructure in our solutions enable resource authorizations and lay the foundation for the needed end-to-end security.