An approach to select and schedule the most appropriate sequence of security patches, for a large set of vulnerabilities in a power generation system, using game theory and reinforcement learning

カテゴリ: 研究会(論文単位)

論文No: CMN24029

グループ名: 【C】電子・情報・システム部門 通信研究会

発行日: 2024/03/25

タイトル(英語): An approach to select and schedule the most appropriate sequence of security patches, for a large set of vulnerabilities in a power generation system, using game theory and reinforcement learning

著者名: Minz Remish Leonard(Hitachi India Pvt. Ltd.),Yadav Geeta(Indian Institute of Technology, Ropar),Kalle Ritesh Kumar(Hitachi India Pvt. Ltd.)

著者名(英語): Remish Leonard Minz(Hitachi India Pvt. Ltd.),Geeta Yadav(Indian Institute of Technology, Ropar),Ritesh Kumar Kalle(Hitachi India Pvt. Ltd.)

キーワード: Power grids|Security|Vulnerability|Countermeasures|Game Theory|Reinforcement Learning|Power grids|Security|Vulnerability|Countermeasures|Game Theory|Reinforcement Learning

要約(日本語): The surge in cyberattacks on power grids emphasizes_x000D_ the need for protecting critical assets and ensuring power_x000D_ reliability. However, maintaining reliability requires promptly_x000D_ addressing vulnerabilities through security measures, which may_x000D_ cause system downtime. Power demand requirements impose_x000D_ strict constraints on grid up-time, creating a trade-off between_x000D_ grid security and power availability. We propose a two stage_x000D_ optimization method to minimize this trade-off for a power_x000D_ generation plant. Firstly, we employ a game theory approach_x000D_ to prioritize the vulnerabilities for applying security countermeasures_x000D_ in the generation plant. Second, we use power demand_x000D_ and supply data of an Indian city to forecast future Opportunity_x000D_ Window and use Reinforcement Learning to efficiently utilize_x000D_ the Opportunity Window to patch the prioritized vulnerabilities_x000D_ from the first stage. We evaluate this approach on supervisory_x000D_ control system of a power generation plant.

要約(英語): The surge in cyberattacks on power grids emphasizes_x000D_ the need for protecting critical assets and ensuring power_x000D_ reliability. However, maintaining reliability requires promptly_x000D_ addressing vulnerabilities through security measures, which may_x000D_ cause system downtime. Power demand requirements impose_x000D_ strict constraints on grid up-time, creating a trade-off between_x000D_ grid security and power availability. We propose a two stage_x000D_ optimization method to minimize this trade-off for a power_x000D_ generation plant. Firstly, we employ a game theory approach_x000D_ to prioritize the vulnerabilities for applying security countermeasures_x000D_ in the generation plant. Second, we use power demand_x000D_ and supply data of an Indian city to forecast future Opportunity_x000D_ Window and use Reinforcement Learning to efficiently utilize_x000D_ the Opportunity Window to patch the prioritized vulnerabilities_x000D_ from the first stage. We evaluate this approach on supervisory_x000D_ control system of a power generation plant.

本誌: 2024年3月28日-2024年3月29日通信研究会

本誌掲載ページ: 67-72 p

原稿種別: 英語

PDFファイルサイズ: 1,345 Kバイト