HIIDS: Hybrid intelligent intrusion detection system empowered with machine learning and metaheuristic algorithms for application in IoT based healthcare

Abstract

This paper presents machine learning and metaheuristic algorithms based hybrid intelligent Intrusion Detection System (HIIDS) for Internet of Things based applications such as healthcare. In IoT based smart healthcare, biomedical sensors sense vital health parameters which are sent to the cloud server for storage and analysis. Health data saved as Electronic Health Record (EHR) is privacy and security sensitive. This work focuses on the detection of security attacks on cloud servers through anomaly based intrusion detection. Popular NSL-kDD dataset containing 41 features with 125,973 samples have been utilized for performance evaluation of proposed HIIDS.To reduce computation cost, metaheuristic algorithms such as Particle Swarm Optimization (PSO), Genetic Algorithm (GA), and Differential Evaluation (DE) are used for best feature selection and supervised learning algorithms such as Known Nearest Neighbor (kNN), Decision Tree (DT) are used for accurate classification of normal and attack class based on selected features. Also a hybrid approach has been presented for feature selection and classification. After dataset pre-processing using python, MATLAB 2019b is used to implement six variants of proposed hybrid algorithms combining GA, PSO, DE with kNN, DT. Performance evaluation has been done based on accuracy, execution time, memory usage and CPU utilization. GA-DT variant gives highest accuracy of 99.88%, 86.40%, 95.39%, 96.90%, 100% of accuracy for DoS, U2R,R2L, Probe and Normal class with the help of 8-10 features compared to other variants such as GA-kNN, PSO-kNN, PSO-DT, DE-kNN, DE-DT. Also outperforms similar state-of-the-art works in terms of classification accuracy, simulation results are given in support. Finally an IoT based healthcare architecture is designed using best performing hybrid GA-DT variant based HIIDS to detect and prevent malicious traffic.