"Electrical Engineering Doctoral Student Mohammadamin Moradi Uses Deep-Q Learning to Find and Combat Power Grid Cybersecurity Weaknesses"
As power grids become more reliant on computer-based systems, they become more vulnerable to cyberattacks. Mohammadamin Moradi, an electrical engineering doctoral student at Arizona State University (ASU), used Artificial Intelligence (AI) to analyze the most damaging attacks against the power grid and best possible defenses with guidance from Ying-Cheng Lai, a Regents Professor of electrical engineering, and Yang Weng, an Assistant Professor of electrical engineering. The US Department of Energy (DOE) and the Israeli Ministry of Energy funded this research through the Israel-US Binational Industrial Research and Development (BIRD) Foundation to help both countries improve their cybersecurity defenses. The researchers used deep-Q Reinforcement Learning (RL), a type of Machine Learning (ML), in conjunction with stochastic game theory, to simulate which cyberattacks would cause the most damage to a power grid and the best countermeasures to keep the grid running as efficiently as possible in the face of such attacks. Deep-Q learning examines the outcomes of inputs in order to maximize the reward for an action. In traditional Q-learning, various user inputs are mapped to output values in a table known as the Q-table. However, creating a Q-table presents numerous challenges because it requires a significant amount of computation as the number of input values grows. When the number of inputs and outputs reaches a certain size, this can cause a computer to struggle and malfunction, prompting Moradi to study deep-Q learning. Moradi also chose deep-Q learning because it can be used in environments with unknown parameters, such as when the optimal attack and defense strategies are unknown before running the deep-Q learning simulation. Although deep-Q learning addresses the issue of required computing power, the algorithm model used by the system to learn must also be optimized to ensure the best results. This is where Moradi got the idea to turn the scenario into a stochastic game. This article continues to discuss the team's study on defending smart electrical power grids against cyberattacks with deep-Q RL.