A Trust Based Framework for Secure Data Aggregation in Wireless Sensor Networks

In unattended and hostile environments, node compromise can become a disastrous threat to wireless sensor networks and introduce uncertainty in the aggregation results. A compromise node often tends to completely reveal its secrets to an adversary which in turn renders purely cryptography-based approaches vulnerable. How to secure the information aggregation process against node compromise attacks and quantify the uncertainty in the aggregation results has become an important research issue. In this talk, we will address this problem by proposing a trust based framework, rooted in statistics and other distinct yet closely coupled techniques. The trustworthiness (reputation) of individual sensor nodes is evaluated with the help of an information theoretic measure, Kullback-Leibler distance, that identifies the compromised nodes through an unsupervised learning algorithm. Upon aggregation, an opinion (metric defining the degree of belief) is generated to represent the uncertainty in the aggregation result.

As the result is disseminated and assembled through the routes to the sink, the opinion will be propagated and regulated by Josang's belief model. Our trust based framework effectively quantifies the uncertainty in the data as well as aggregation results. Experimental results demonstrate that this framework provides a powerful mechanism for detecting compromised nodes and reasoning about uncertainty in the network. It can also purge false data to accomplish robust aggregation in the presence of compromised nodes.