Wöhnert, Kai HendrikWendzel, SteffenWressnegger, ChristianHartmann, LauraFreiling, FelixArmknecht, FrederikReinfelder, Lena2024-04-192024-04-192024978-3-88579-739-5https://dl.gi.de/handle/20.500.12116/43961The electrical grid is transitioning towards a decentralized structure, spurred by the inclusion of renewable energy. This paper addresses the complex security challenges faced due to its decentral network architecture. Traditional network security methods are insufficient in safeguarding against threats in this evolving environment. The focus is the creation of a decentralized intrusion detection system (IDS) using a machine learning approach optimized for resource-constrained devices. Preliminary evaluations indicate that smaller recurrent neural networks can effectively detect denial of service attacks in simulated networks. Future work will involve real-world data analysis and field tests in genuine smart grid environments.enintrusion detectionmachine learningsmart gridText/Conference Paper10.18420/sicherheit2024_0201617-5468