Time series dataset for network security situational awareness.

Area of Science:

• Cybersecurity
• Network Security
• Data Science

Background:

• Existing network security situational awareness (NSSA) datasets are often outdated, small, or inaccessible.
• This limits the development and validation of advanced machine learning models in cybersecurity.
• There is a critical need for large, usable datasets for NSSA research.

Purpose of the Study:

• To introduce a novel, comprehensive dataset for network security situational awareness (NSSA).
• To provide a resource that overcomes the limitations of previously available cybersecurity datasets.
• To support the training and evaluation of neural networks and other machine learning models.

Main Methods:

• Collected approximately three billion cybersecurity alerts from the Warden system between 2017-2018 and 2023-2024.
• Processed raw alerts into time-series data, suitable for machine learning applications.
• Utilized a 'silver standard' for data labeling, where alerts are labeled by the detection system itself.

Main Results:

• A substantial, multi-part dataset comprising time-series data from billions of cybersecurity alerts.
• The dataset spans two distinct time periods, capturing evolving threat landscapes.
• The data is derived from a well-documented and substantial source, ensuring usability for research.

Conclusions:

• The presented dataset offers a valuable resource for the cybersecurity research community.
• It enables the development of more robust and accurate NSSA systems, particularly those using deep learning.
• This initiative aims to advance the field of network security by providing accessible, large-scale data.