Applying Deep Reinforcement Learning for Detection of Internet-of- Things Cyber Attacks
Curtis Rookard, Anahita Khojandi
Abstract
As society becomes more interconnected, smaller computing platforms such as embedded systems and internet-of-things (IoT) devices have become increasingly common. Un-fortunately, these computing platforms are still subject to cyber attacks. The usage of network intrusion detection systems is an established approach for the detection of cyber threats. In this study, we present a reinforcement learning-based network intrusion detection system to detect attacks on IoT systems using the TON-IoT dataset. Specifically, we employ the usage of a deep Q-network (DQN) for cyber threat detection. Our reinforcement learning model is compared against other popular machine learning models. We find that our DQN performs the best for cyber attack detection.