Wireless sensors networks(WSNs) consists of large number of tiny sensors with the capabilities of sensing, processing and communication, a vital part of today many incredible applications that mostly including monitoring and tracking activities. In order to carry out these applications successfully, the Wireless sensors networks (WSNs) can be seen in many different forms ranges from small to large scale distributed wireless sensor networks in hostile environments. Recently, the wireless sensor networks (WSNs) have become a focus of intensive research especially for efficient utilization of limited resources of nodes and careful design of these networks in order to get better results for WSNs oriented applications.
One of such attempt from research community can be seen in the form of hierarchical structuring or clustering algorithms for this large scale distributed wireless sensor networks for efficient utilization of limited resources at nodes and overcoming the various challenges that results from it. Thus, the cluster based wireless sensor network (CBWSNs) have the capability to elevate the related problems of resource limitation of nodes while performing their basic activities i.e. routing, aggregating and forwarding. On the other hand, the security is become a great concern, when the nodes inside the cluster compromised and start misbehaving selfish in order to save their limited resources for their own use. Such attacks are known as internal attacks or passive attacks. At present, the existing security measures for WSNs cannot ensure that these problems will not be launched. Therefore, it is important to protect the CBWSNs from internal attacks, which is the main goal of this thesis.
This thesis scrutinizes the security problems of CBWSNs and proposes relevant solutions.
The main contributions of this thesis to mitigate the selfishness oriented anomalies of nodes in the cluster are summarized below.
Initially, this thesis developed a security framework based on novel clustering algorithm and utilizing primary reputation system evaluation for finding selfishness problems at clusters. The novelty of clustering process results in the clusters with three basic nodes i.e. Head (CH) node, Inspector Node (IN) and Member Nodes (MNs). The selection criteria of these nodes are to utilizing the reputation system evaluations at each node before and after cluster formation. After cluster formation inspector node overhears cluster head node transmission and updates its reputation status and special way of working of all three types of nodes can further strengthen the application of the proposed security framework. In this way, these reputation systems at nodes will keep track of the neighbor nodes behavior and force them to be cooperative.
Secondly, the improved security framework based clustering algorithm utilizing residual energy and reputation data of nodes for finding selfishness free cluster head node for consistent clusters. The novelty of security framework is now due to use of reputation system based on data mining i.e. Bayesian rule at each node and Support Vector Machine scheme at base station for analyzing reputation data for detecting selfishness at nodes. Further, considering remaining energy can improve the consistency of clusters and make them free of attacks. Thus, after clustering the whole network will be divided into groups and having three basic nodes i.e. Custer Head (CH) node, Member Nodes (MNs) and Sink Node (IN).
This work utilized the fourth generation programming language MATLAB for simulations. The simulation results show that the security framework for the detection of the internal attacks is effective and valuable.