Black Hole Attack Detection in Healthcare Wireless Sensor Networks Using Independent Component Analysis Machine Learning Technique

Abstract

Wireless Sensor Networks (WSNs) are self-configured infrastructure-less networks are comprising of a number of sensing devices used to monitor physical or environmental quantities such as temperature, sound, vibration, pressure, motion etc. They collectively transmit data through the network to a sink where it is observed and analyzed. The major issues in WSN are interference, delay and attacks that degrade their performance due to their distributed nature and operation. Timely detection of attacks is imperative for various real time applications like healthcare, military etc. To improve the Black hole attack detection in WSN, Projected Independent Component Analysis (PICA) technique is proposed herewith, which detects black hole attack by analyzing collected physiological data from biomedical sensors. The PICA technique performs attack detection through Mutual information to measure the dependence in the joint distribution. The dependence among the nodes is identified based on the independent probability distribution functions and mutual probability function. The black hole attack isolation is then performed through the distribution of the attack separation message. This supports to improve Packet Delivery Ratio (PDR) with minimum delay. The simulation is carried out based on parameters such as black hole attack detection rate (BHADR), Black Hole Attack Detection Time (BHADT), False Positive Rate (FPR), PDR and delay.