A Deep Reinforcement Learning Method for Accurate and Efficient Anomaly Detection

Anomaly detection is a crucial service in contemporary data networks, and the development of new fast and robust algorithms capable of detecting and classifying dangerous traffic is essential to cope with evolving threats and increasing detection difficulty. In this paper, we present a novel anomaly detection algorithm that uses deep reinforcement learning (DRL) to efficiently learn and adapt to anomalies in time series data. Compared to alternative DRL methods using in anomaly detection, the approach posited in this paper presents various advantages. Primarily, it outperforms other techniques in scenarios with continuous action spaces. Additionally, it contributes to the stabilization of the learning procedure and enhances the convergence of the algorithm. Finally, it facilitates the acquisition of superior policies and circumvents the risk of being trapped in local optima.