Research On Intrusion Detection Of Subtractive Clustering Based Placement In Wireless Sensor Networks

Wireless sensor networks developed as a new research field rapidly and can beapplied in fields of military, environment, medical health, household, spaceexploration, disaster relief, etc potentially. However, the security of wireless sensornetworks is in relation to the security of country and social from personal privacy tostate secrets. Intrusion detection is important security techniques for wireless sensornetworks. Intrusion detection system collects and analyzes information of networksthrough some key points to find the behavior of violating security strategy and signsof being invaded.To solve the security problem of wireless sensor networks, the intrusiondetection techniques of subtractive clustering based placement is presented in thispaper. The intrusion detection modules are distributed through the subtractionclustering algorithm. Intrusion detection modules are embedded in all sensor nodes,and subtractive clustering algorithm is used to classify the sensor nodes. Then clusterhead nodes activate intrusion detection modules, which are embedded in cluster headnodes for the neighborhood detection. After the regular period，a new cluster headnode will be selected for detecting according to the residual energy and other factorsin the original cluster, and the former cluster head stop monitoring function. This cansatisfy the requirements of intrusion detection rate, guarantee the balance of energyconsumption of the detection nodes, and make the detection nodes detect the entirenetwork more effectively. A kind of intrusion detection scheme based onspecification was proposed to detect the attacks in wireless sensor networks. In thescheme, threshold was set for characteristic quantity of neighborhood nodes in unittime according to the theory of probability. The method of threshold setting wasversatile, and threshold could update adaptively to accord with the properties ofsensor nodes changing over time. Within the scope of threshold value, the networknode behavior is normal. Meanwhile behavior of abnormal nodes is not in thethreshold value.The NS2simulation software was used to simulate wireless sensor networks,and the proposed intrusion detection based on subtraction clustering was realized. The simulation experiment verifies that the proposed subtractive clustering basedplacement strategy of intrusion detection modules is better than other schemes, andthe proposed scheme based on specification for intrusion detection has betterdetection effect. In this paper the proposed subtraction clustering based intrusiondetection techniques not only can meet the request of high detection rates and lowfalse positives, but also can ensure detection node has good energy efficiency.