Search

Wireless Body Area Network (WBAN) has become a reality where patient health conditions can be monitored remotely. Lightweight small sensor devices gather the health parameters and transmit the data to a remotely located medical facility. These kinds of resource-constrained networks are prone to several routing attacks, which can disrupt the whole network. Cryptographic approaches are inefficient in fighting against routing attacks, while Machine Learning based Intrusion Detection System (IDS) can identify typical and uncommon security attacks. For this purpose, a dataset has a significant role to play. This paper focuses on building a specialized dataset and an accurate Intrusion Detection Model for WBAN. A simulation has been performed using E-HARP routing protocol to simulate four routing attacks, such as Wormhole, Blackhole, Byzantine, and Scheduling attack, in addition to regular data transmission (without attack). Deer Hunting Optimization Algorithm (DHOA) and Adaptive Neuro Fuzzy inference System (ANFIS) have been merged to model a highly accurate Intrusion Detection Model. A specialized dataset has been used to train and test the hybrid DHOA-ANFIS model. The proposed model has been evaluated and compared with similar hybrid models such as CSO-ANFIS, PSO-ANFIS, BA-ANFIS, GA-ANFIS, and ABC-ANFIS. The proposed model achieves the highest overall accuracy in experiments, demonstrating its superiority over competing models. It has obtained an overall accuracy of 97.70%, a Detection Rate of 97.03%, a Precision of 96.20%, and the lowest False Alarm Rate of 0.90%. The Proposed model also outperforms state-of-the-art related works on designing machine learning based IDS for WBAN and IoT based healthcare. [ABSTRACT FROM AUTHOR]

Health monitoring using Body Area Wireless Sensor Network (BA-WSN) has gained immense popularity due to usability, ubiquitous support, and real-time performance. It is a special kind of Wireless Sensor Network (WSN) that spans over the human body. Although BA-WSN is very useful but it may suffer from security and privacy issues due to compromised sensor nodes by intruders. To design a secure BA-WSN based health monitoring system, it is required to filter out malicious data packets generated by the compromised nodes. An intrusion Detection System (IDS) can be utilized for this purpose. This paper presents a Machine Learning based Intrusion Detection System (MLIDS) for BA-WSN based health monitoring framework. Specialized dataset WSN-DS has been used to train the intrusion detection model. Dataset contains four security attacks such as Blackhole attack, Grayhole attack, Scheduling attack, Flooding attack data as well as normal data packets which are simulated using Network Simulator-2. Five well-known classification algorithms such as Random Forest, kNN, SVM, J48, and Naive Bayes have been applied for the selection and generation of the best model in terms of detection accuracy. Experimental results prove that Random Forest based Intrusion Detection Model has the highest classification accuracy of 99.67%, 98.7%, 92.7%, 98.9%, 99.9% for Blackhole attack, Flooding attack, Scheduling attack, Grayhole attack as well normal packet respectively. Experimental results also show that our achieved results outperform relevant work in terms of accuracy. [ABSTRACT FROM AUTHOR]