Attack Detection and Classification of Heterogeneous Wireless Sensors Using Co-clustering

In a Wireless Sensor Network a large number of sensors are deployed for the purpose of sensing data and then to bring the data back securely to nearby base stations. The base stations then perform the costly computation on behalf of the sensors to analyze the data sensed by the sensors. Due to resource limitations of the nodes and also due to the vulnerability of physical captures of the nodes, the traditional cryptographic techniques are very complex and should not fit for energy constrained environments. Data mining techniques can be applied to find the malicious behavior of the nodes of the Sensor Network. By analyzing the traffic patterns one can differentiate the normal behavior from malicious behaviour.Those techniques are used to identify various attacks. This paper addresses the issue of Attacks using data mining techniques. There exist two types of attacks: (i) External and (ii) Internal. External attacks are those in which an attacker manipulates the communication between pairs of trusted nodes and causes the nodes to de-synchronize. Internal attacks are those in which internal attackers report false clock references to their neighboring nodes proposed an approach to develop a protocol. The protocol not only finds malicious node(s) but also counts them within the group using data mining clustering techniques.