一种基于异常检测的传感器网络容侵数据聚合方案

Anomaly detection based intrusion-tolerant data aggregation scheme in sensor networks

针对传感器网络数据聚合协议易受恶意数据注入攻击的问题,文中采用层次化贝叶斯空时模型表征聚合通信模式下感知数据的统计特性,并基于该模型提出了一种基于差分滤波的容侵数据聚合机制,在数据聚合过程中消除由恶意节点注入的错误数据。文中分别通过理论分析和基于Tiny OS平台的仿真实验评估了方案的执行效率和有效性,结果均表明,文中提出的方案能够有效抵御传感器网络数据聚合协议中的错误数据注入攻击。

Data aggregation is known to be vulnerable to false data injection launched by compromised sensor nodes. To address this problem,a hierarchical Bayesian spatial-temporal （HBST） methodology is used to capture the statistical characteristics of sensory data in the aggregation-based communication mode, and an intrusion-tolerant aggregation scheme is proposed in which a false data detection technique based on divided difference filtering （DDF） is used to eliminate the false data injected by compromised sensor nodes. The performance of the proposed scheme is evaluated by the theoretical analysis and experi- ments on the TinyOS platform. Theoretical and experimental results indicate that the proposed scheme is suitable to resist false data injection attacks for the aggregation in wireless sensor networks.