CYBERATTACK DETECTION AND MITIGATION IN POWER DISTRIBUTION SYSTEMS USING MACHINE LEARNING
Keywords:
Intrusion detection, cyber security, anomaly detection, q-learning, reinforcement learning, multi-agent systems, distribution automationAbstract
In order to safeguard power delivery systems, cyber-physical system security is essential. The switches in the distribution network that can be remotely managed will be the target of the direct switching attacks that are being planned. Voltage issues and power outages may result from certain designs, which operate predominantly in a radial direction. Modern optimization methods are believed to be able to determine the assailant's settings by observing the manner in which the attacker communicates with the power system user (defender). As a result, they experience a decline in their appeal. It is becoming increasingly common to employ information from both centralized and decentralized security systems to identify intrusions that were premeditated. This may result in an issue with one of the components. As a result, new mathematical models are developed for both the adversary and the friend. The models compare the discovered assaults in a decentralized manner to determine the attacker's objectives, without the assumption that they are aware of the attacker's system set-up.
