Clique-based group key assignment in Wireless Sensor Networks

Security has become the corner stone of research in wireless sensor networks (WSNs). Due to the unique operational environment of WSNs, where the communication medium is open to eavesdroppers, the threats manifest in new forms necessitating the safeguarding of communication. Consequently, key management protocols have become paramount in mitigating the damage caused. Numerous group-key protocols have been proposed in an effort to secure both inter and intra network communication. However, the group-key protocols in literature neither address the issue of the size of a group nor its geographic boundaries. Therefore, in applications like watchdogbased reputation monitoring systems, malicious users are encouraged to pollute the reputation values by bad-mouthing benign nodes and false-praising other collaborating malicious nodes. On the otherhand, though pairwise-key protocols can overcome the above drawback of group-key protocols, they are highly restrictive and impose substantial storage overhead on resource constrained sensors. They do not suit the reputation monitoring systems either, since messages encrypted with pairwise-keys render promiscuous watchdog monitoring systems useless. In this paper, we propose CAGE a novel, distributed, clique-based group-key assignment protocol, which distinctly addresses the size and geographic restrictions on groups. Our protocol is a simple distributed method, yet effective in securing the neighborhood communication and ideally suits the promiscuous monitoring requirements of reputation and trust-based systems. CAGE ensures pairwise connectivity of all group members which thwarts information asymmetry (bad-mouthing) attacks. We confirm through simulations and analysis that CAGE strikes an optimal balance between pairwise-key and group-key protocols and that it achieves an optimal solution. We then present the improvised CAGE, which we refer to as the Extended CAGE (E-CAGE). E-CAGE achieves the same results as CAGE with mitigated computation time compared to CAGE. We confirm this through simulation results.