Authors:
Emilie Frost
1
;
2
;
Julia Heiken
1
;
2
;
Martin Tröschel
2
and
Astrid Nieße
1
;
2
Affiliations:
1
Digitalized Energy Systems, Carl von Ossietzky Universität Oldenburg, Ammerländer Heerstraße 114-118, Oldenburg, Germany
;
2
Distributed Artificial Intelligence, OFFIS e.V., Escherweg 2, Oldenburg, Germany
Keyword(s):
Anomaly Detection, Self-Organizing Systems, Multi-Agent Systems.
Abstract:
In Cyber-Physical Energy Systems (CPES), multi-agent systems are expected to perform a variety of tasks. The increase in digital interconnections and distributed structures in CPES leads to more cyber access points, which increases the risk of cyber attacks. The effect of a manipulated or corrupted agent, as caused by cyber attacks, on the communication of an agent system is investigated in this paper. Anomaly detection is an important prerequisite to identify and mitigate malicious behavior and thus protect the critical infrastructure of CPES. Since in distributed systems, some information is only available in a distributed way, this paper introduces a centralized and a distributed architecture for anomaly detection. For this, a dataset is presented from an agent-based energy system control use case, including anomalies in agent behavior.