Quantum Entropy and Reinforcement Learning for Distributed Denial of Service Attack Detection in Smart Grid

Distributed Denial of Service (DDoS) threats in Smart Grid are very challenging and considered one of the most destructive cyber-attacks. They are harmfully affecting the power sector and causing substantial financial loss. Classical Reinforcement Learning (RL) models are proposed as a promising solution to the DDoS problem as they learn optimal mitigation policies under different attack scenarios. However, as the RL environment is generally dynamic, which leads to more computation capabilities, classical RL efficiency can be limited as a result of serial processing using classical computing. In this paper, a Quantum Entropy Q-Learning (QEQ) is proposed to fight DDoS in Smart Grid. The proposed framework is compared to a classical Q-Learning (QL) model with and without the Entropy method. The convergence speed and the total rewards performed with the same conditions for QL and are Entropy Q-Learning (EQ) are analyzed to show the QEQ performance. Moreover, the accuracy, precision, recall, and F1 score are evaluated to prove the effectiveness of our QEQ in fighting DDoS attacks. Using the Canadian Institute for Cybersecurity Intrusion Detection System (CICIDS 2019) dataset, a thorough systematic simulation using MATLAB, Python, the open-source Qiskit software, and the Harrow-Hassidim-Lloyd (HHL) quantum algorithm is carried out. This proposed QEQ is better performing the DDoS detection as it is faster enough and more adaptable to dynamic environment changing and able to improve the agent’s decision-making in a changing state action space. Finally, conclusions with some open issues of the Quantum Q-Learning are presented.