ADSBAN: Anomaly detection system for body area networks utilizing IoT and machine learning

Abstract

SummaryBody area networks (BANs) play a pivotal role in modern healthcare, enabling real‐time data collection and monitoring of vital patient parameters, thereby empowering healthcare providers to respond swiftly to any serious health conditions. These networks depend on strategically placed sensors on the patient's body to collect important health data. The integrity of this data is really important, as problems with a component can cause doctors to make mistakes that could be life‐threatening. Consequently, developing and evaluating robust anomaly detection methods for BANs is paramount. To address this concern, an anomaly detection system for body area networks (ADSBAN) has been proposed in this article. Different machine learning methods have been tested and implemented to improve the proposed system's performance. These include decision tree, K‐nearest neighbor, logistic regression (LR), random forest, AdaBoost (Adaptive Boosting), and XGBoost (Extreme Gradient Boosting). Further, this framework has been tested with emulated and standard datasets and compared with existing methodologies. The research employs an IoT Flock emulator to replicate BAN conditions and simulate two distinct attack scenarios. Wireshark aids in thoroughly analyzing network traffic, while Python, in conjunction with tools like Keras and Pandas, is instrumental in implementing the ML models. Further, a CIC flowmeter has been utilized to convert the .pcap file format into .CSV file format. Comprehensive evaluation metrics such as precision, recall, accuracy, F1 score, and the Mathew correlation coefficient (MCC) demonstrate the consistent superiority of LR on emulated (i.e., traffic generated through a simulator IoT Flock) and standard datasets (BoT‐IoT). LR shows the highest accuracy of 99.92% on emulated and standard datasets, while XGBoost has the highest average accuracy of 99.92% on standard dataset. This research significantly bolsters the reliability of healthcare data, instilling confidence among healthcare professionals in their decision‐making processes. By safeguarding the integrity of vital health data in BANs, it advances the quality of patient care. It underscores the indispensable role of ML techniques in fortifying the resilience of healthcare systems.