ICSCASD-MPLC Dataset

This study investigates cybersecurity issues in In- dustrial Control Systems, focusing on Programmable Logic Con- trollers within a custom-built network environment. To address the need for realistic datasets in Industrial Control Systems security research, we created a setup using industrial switches that enables real-time execution of various cyber-attacks, includ- ing Denial-of-Service, Man-in-the-Middle, Address Resolution Protocol Spoofing, Data Injection, and Reconnaissance. Data flow to the Programmable Logic Controllers was managed through computers running Node-RED, while a Kali Linux system was used to execute the attack scenarios. A separate system equipped with Wireshark and Python scripts captured and analyzed both legitimate and malicious traffic. This process generated a detailed dataset, which was used to train machine learning models, specifically Decision Trees and eXtreme Gradient Boosting. The Decision Tree model achieved 99% accuracy in binary classi- fication and 97% in multi-class tasks, while eXtreme Gradient Boosting performed slightly better, reaching 99% accuracy in binary classification and 98% in multi-classification. This work contributes to the field of ICS cybersecurity by revealing vul- nerabilities in industrial networks and emphasizing the need for robust security protocols to protect critical infrastructure. By providing a replicable approach for simulating cyber threats, this research supports the development of machine learning-based intrusion detection systems, reinforcing the overall resilience of industrial control systems.