Context centric cluster computing in ad hoc network (C4)

Abstract

Cluster computing in ad hoc network draws special attention from the research community due to its inherent support for group communication in varied application domains like emergency and rescue operations, etc. Context sensitive aware cluster computing yields maximum benefits and advantages than the traditional computing. The Cluster Head (CH) and the Cluster Members (CM) effectively coordinate the cluster communication through robust key and trust management. The key management is a crucial and resource demanding phase that has to be effectively monitored by CH. Several factors necessitate the mobility of CM across clusters like node leaving/joining the parent/foreign clusters, the corrupted CH forbidding its current CM to join a foreign cluster. The compromise of CM and periodic key refreshment/update are the other factors that incite the rekeying mechanism needed to preserve the integrity and credibility of the cluster communication. The rekeying kinematics has to be maintained at a low profile to overcome the storage, communication and computational overhead associated with it. The Drifted Existing Cluster Member (DECM) rejoining/re-affiliating to the cluster to which it was previously associated exempts strict divorce/separation from the existing cluster unless it is appended to Node Corruption List (NCL). This process of DECM rejoining the parent cluster demands partial rekeying mechanism enabled through context mining for better resource optimization. The CH maintains collective key instance scenarios/contexts that are rendered to Rejoined DECM (RDECM) for ensuring minimal rekeying rate to preserve forward and backward secrecy. The rejoined DECM after stiff authentication and authorization check by CH seamlessly synchronizes with the cluster through its self-learning ability of cluster contextual mining (CCM) to get introduced to the recently added CM and re-establishing the point of contact with its old neighbors. This process paves way for relaxing the constraints on maintaining stiff forward and backward secrecy in cluster/group computing. This approach precludes the RDECM from having complete history of the cluster to which it reassociates by acquiring glimpse of the several scenarios/context captured by CH. This re-affiliation/reunion entailing minimum change overhead has to be propagated to other CH that treats the new node as its good old Samaritan incumbent node. Various graphs are simulated to adjudge the performance escalation in cluster computing through adoption of this novel technique.