A dataset collected in real-world industrial control systems for network attack detection.

Area of Science:

• Cybersecurity
• Industrial Control Systems (ICS)
• Industry 4.0

Background:

• Industrial Control Systems (ICS) are increasingly open in Industry 4.0, raising cybersecurity risks.
• Existing ICS datasets lack diversity, realism, and scope for effective intrusion detection.
• There is a critical need for high-quality datasets to train robust ICS intrusion detection systems.

Purpose of the Study:

• Introduce the ICS Network Attack Dataset (ICS-NAD) for realistic ICS cybersecurity research.
• Provide a comprehensive dataset covering diverse ICS attack types and real-world scenarios.
• Facilitate the development and validation of advanced ICS intrusion detection methods.

Main Methods:

• Collected data from real-world Industrial Control Systems (ICS) environments.
• Included two attack traffic patterns and 20 common ICS attack types.
• Extracted and labeled 60 features from raw network traffic (PCAP) and provided CSV files.

Main Results:

• The ICS-NAD dataset contains 245.96 GB of data, including raw traffic and labeled features.
• Validated dataset utility with 10 machine learning and deep learning classification models.
• Demonstrated the dataset's suitability for training and testing intrusion detection systems.

Conclusions:

• ICS-NAD addresses the limitations of existing datasets, offering a valuable resource for ICS cybersecurity.
• The dataset supports academic and engineering research in Industrial Control Systems (ICS) network attack detection.
• Public availability of ICS-NAD promotes advancements in securing Industry 4.0 infrastructure.