Recovering an OSPF Network from Malicious Attacks: An Experimental Evaluation of Recovery Techniques

The routing infrastructure of the Internet is a high-value target for malicious actors. The Open Shortest Path First (OSPF) protocol is one of the most widely used protocols for Autonomous System (AS) internal routing, and has been the subject of attacks targeted at partitioning the network. Because of the widespread existing penetration of OSPF, research in defending against such attacks has focused toward prevention and detection techniques working in tandem with OSPF routers, rather than attempted modification of the protocol itself. However, comparatively few studies have attempted to address the topic of a network capable of recovery, which signifies a system that can be proof to an attack without the necessity for any specific attack to be explicitly detected, localized, or countered. In this paper, we first examine these existing recovery techniques, and come up with the underlying common mechanisms that they utilize. Then we examine the continued usefulness of these approaches with the more modern persistent OSPF attacks, by running actual attacks against an isolated network formed of typical commercial network elements, while allowing each such recovery mechanism to defend the network. Our experiments show that three of the five basic mechanisms can no longer defend against partitioning attacks when attacks are persistent, and provides relevant performance results for the other two. Our results also point the way to further improving these mechanisms for even more sophisticated attacks, in the future.