Home // SECURWARE 2025, The Nineteenth International Conference on Emerging Security Information, Systems and Technologies // View article
Authors:
Robert Altschaffel
Stefan Kiltz
Jana Dittmann
Tom Neubert
Laura Buxhoidt
Claus Vielhauer
Matthias Lange
Rüdiger Mecke
Keywords: SCADA; hidden-non-malicious-dummy; cyber-security.
Abstract:
Cyber-Security in Industrial Control Systems (ICS) is a topic of growing relevance. Attack scenarios include the exfiltration of critical process data, the infiltration of commands and the manipulation of the controlled physical processes. Machine learning based detection mechanism are employed against these attacks. However, such machine learning based approaches rely on training data. This paper addresses two core challenges with regards to such machine learning approaches: 1. the required training data containing such attacks is usually difficult to obtain and 2. information about the detection rates is necessary in order to deploy the mechanisms for detection in a fashion benefiting security incident management. As such, this paper discusses an approach to generate such training data containing hidden non- malicious-dummy data representing attacks for five different attack scenarios, means to ensure that these dummies do not negatively affect the system under test, different strategies for injection and detection. This synthetically generated facility-specific data is then used for evaluating the usefulness of such machine learning detection approaches in ICS security management.
Pages: 69 to 75
Copyright: Copyright (c) IARIA, 2025
Publication date: October 26, 2025
Published in: conference
ISSN: 2162-2116
ISBN: 978-1-68558-306-4
Location: Barcelona, Spain
Dates: from October 26, 2025 to October 30, 2025