Sphinx

Abstract

Nowadays structured overlay networks are used in controlled environments like company wide area networks. Even though these structured overlay networks are usually closed, there is still the risk that one or more nodes get compromised by attackers. When these networks provide business critical services, serious harm can be done by a compromised node. Because these networks are company controlled, properties like available bandwidth and latency times are much better known than in a general wide area network like the Internet. This knowledge can be used to create a system where malicious nodes in the network can be detected and located in a relatively short period of time and with little overhead. This thesis describes the design and implementation of the Sphinx protocol that provides this functionality. We show that it is possible to detect and locate malicious nodes in a distributed hash table in a short period of time, with little overhead, and with high sensitivity. We are able to detect and locate the nodes that cause latency deviations, dropping of messages, mis-routing of messages, and changing of the message payload.