The effect of time dimension and network dynamics on key distribution in wireless sensor networks

Official URL: http://192.168.1.20/record=b1276348 (Table of Contents)

The majority of studies on security in resource limited wireless sensor networks (WSN) focus on finding an efficient balance among energy consumption, computational speed and memory usage. Besides these resources, time, network dynamics (e.g. routing), and implementation and integration issues of the security solutions are relatively immature aspects that can be considered in system design and performance evaluations. In the first part of this thesis, we develop and analyze different implementation options of a Random Key Predistribution scheme in a real network simulation environment. Implementation options include Proactive Key Establishment and Reactive Key Establishment. In Proactive Key Establishment, pairwise keys are established at the beginning, prior to start of application. In Reactive Key Establishment, keys are established only whenever needed by the application during its execution. In literature the latter is known to preserve energy since it reduces useless key establishments; however, it also introduces delay in application traffic. We implement the reactive key establishment in such a way that key establishment traffic and energy consumption are reduced. As a result our reactive key establishment implementation has similar throughput performance with proactive scenarios despite the longer lifetime of reactive scenario. We also simulate an attack scenario and measure different metrics including a novel one. This new metric, the packet compromise ratio, reflects the harm caused by the adversary in a more realistic way. In our simulations, we show that packet compromise ratios are very high as compared to link compromise ratios for a long period. However, when the majority of nodes die, link compromise ratios exceed packet compromise ratios. This is an indication to the fact that link compromise ratios seem high even though there is no high amount of traffic in network to be compromised by adversary. Due to the results showing that classical key distribution schemes in WSNs have actually low resiliency, in the second part of this thesis, we propose new deployment models that improve resiliency. In a recent study by Castelluccia and Spognardi, the time dimension is used to lower the ratio of compromised links, thus, improving resiliency in key distribution in WSNs. This is achieved by making the old and possibly compromised keys useful only for a limited amount of time. In this way, the effect of compromised keys diminishes in time, so the WSN selfheals. We further manipulate the time dimension and propose a deployment model that speeds up the resiliency improvement process with a tradeo between connectivity and resiliency. In our method, self healing speeds up by introducing nodes that belong to future generations in the time scale. In this way, the duration that the adversary can make use of compromised keys becomes smaller.