Designing security systems for wireless sensor networks presents a challenge due to their relatively low computational resources. This has rendered many traditional defense mechanisms based on cryptography infeasible for deployment on such networks. Reputation and anomaly detection systems have been implemented as viable alternatives, but existing implementations still struggle with providing efficient security without a significant impact on energy consumption. To address this trade-off between resource consumption and resiliency, we designed TrustSense, a reputation management protocol for clustered WSNs. It is a semi-centralized family of algorithms that combine periodic trust updates, spatial correlation, and packet sequence validation at the cluster-heads' hierarchy to relieve the sensor nodes of unnecessary opinion queries and trust evaluation computation. We compared the efficiency of TrustSense with legacy reputation systems such as EigenTrust and the results of simulations show a significant improvement in reliability and energy usage while maintaining an acceptable path length with varying numbers of malicious nodes. We believe the approach of combining different techniques from various classes of intrusion detection systems unlocks several possibilities of achieving better results by more complex and versatile composition of these techniques.
翻译:由于计算资源较少,设计无线传感器网络的安全系统是一项挑战。这使得许多基于加密的传统防御机制无法在这类网络上部署。光荣和异常探测系统已经作为可行的替代办法实施,但现有的实施仍然在努力提供高效安全,而不会对能源消耗产生重大影响。为了解决资源消耗和复原力之间的这种权衡,我们设计了TustSense,这是一组集成的WSNS网络的声誉管理协议。这是一个半集中的算法组合,它将定期更新信任、空间相关性和集成序列验证结合起来,在集束头头的层级上减少不必要的意见查询和信任评估计算等传感器节点。我们比较了信任系统与EigenTrust等传统名声系统的效率和模拟结果,显示可靠性和能源使用有显著改进,同时保持一个可接受的路径长度,同时有不同数量的恶意节点。我们认为,将不同类别的入侵探测系统的不同技术结合起来,通过这些技术的更复杂和多用途构成,可以取得更好的结果。