A Deep Reinforcement Learning Based Intrusion Detection System (DRL-IDS) for Securing Wireless Sensor Networks and Internet of Things

Abstract

Many modern infrastructures incorporate a number of sensors and actuators interconnected via wireless links using Wireless Sensor Network (WSN) and Internet of Things (IoT) technology. With a number of mission-critical infrastructures embracing these technologies, the security of such infrastructures assumes paramount importance. A motivated malicious adversary, if not kept in check by a strong defense, can cause much damage in such settings by taking actions that compromise the availability, integrity, confidentiality of network services as well as the privacy of users. This motivates the development of a strong Intrusion Detection System (IDS). In this paper, we have proposed a new Deep Reinforcement Learning (DRL)-based IDS for WSNs and IoTs that uses the formalism of Markov decision process (MDP) to improve the IDS decision performance. To evaluate the performance of our scheme, we compare our scheme against the baseline benchmark of standard reinforcement learning (RL) and the supervised algorithm of machine learning K-Nearest Neighbors (KNN). Through our a thorough simulation-based performance analysis, we demonstrate that our model DRL-IDS returns superior performance in terms of improved detection rate and enhancement the production of accuracy with reduced number of false alarms compared with this current approaches.