Robust IoT Security: Intrusion Detection Leveraging Contributions from Multiple Systems

Purpose and goal:

Intrusion Detection Systems are critical components of an effective Internet of Things (IoT) cybersecurity defense strategy. Effective solutions based on machine learning (ML) relies on the availability of data about earlier attacks. The objective of the project is to explore and provide privacy-preserving and robust techniques to build strong IDS with contributions from multiple actors and systems.

Specifically, the project focuses on methods for mitigating the expected data heterogeneity in federated learning among a set of IoT network providers. Expected results and effects:

The expected outcome of the project are novel methods based on federated learning (FL) specifically tailored for heterogeneous data from different actors and systems, and new platform enablers implemented in an open-source FL system, with future deployments across Swedish industry. Approach and implementation:

The project is a collaboration between Uppsala University (information technology) and Scaleout Systems AB. Selected results from research on knowledge sharing under data heterogeneity and model robustness will be implemented in Scaleout´s open-source federated learning platform, and thereby be made accessible for researchers and industry.