Harmony search Hawks optimization-based Deep reinforcement learning for intrusion detection in IoT using nonnegative matrix factorization
Abstract
With the advent of the Internet of Things (IoT), the security of the network layer in IoT is getting more attention in recent decades. Various intrusion detection methods were developed in the existing research works, but the capability to detect malicious and intrusion activities in the complex Internet environment poses a challenging task in IoT. Hence, an effective and optimal intrusion detection mechanism, named Harmony Search Hawks Optimization-based Deep Reinforcement Learning (HSHO-based Deep RL), is proposed in this research to detect malicious network activities. The proposed Harmony Search Hawks Optimization (HSHO) algorithm is designed by integrating Harmony Search (HS) with the Harris Hawks Optimization (HHO) algorithm. However, the optimal detection result that is effectively achieved through the fitness measures such that the minimum fitness value is only declared as the optimal solution. The Deep Reinforcement Learning (Deep RL) classifier effectively detects the malicious or intruder behaviors and generates a satisfactory result. By reducing the dimensionality of data using nonnegative matrix factorization, the data is optimally fit to perform intrusion detection process in the IoT environment. The proposed HSHO-based Deep RL obtained better performance in terms of the metrics like accuracy (96.925%), True Positive Rate (TPR; 96.90%), and True Negative Rate (TNR; 97.920%) with respect to KK-fold.
