Building Redundancy and Cyber-Resiliency into Intelligent Control Systems by design using Federated Learning.
The cyber security domain is an area that benefits immensely from collaboration. Parties can work together to address mutual threats, such as intelligent anomaly detection across private networks without sharing the data in these networks. Whether prompted by ongoing wars on our doorstep that increase the frequency of attacks or the growing rate of ransomware attacks, just as a single example the shutdown of all IT systems in Irish Nation Health Service in May 2021. There’s a common goal to build cyber-resiliency to protect the data we’re all chasing to gather. This effort can be seen in the European Union’s strategy on protecting the open internet and stricter laws governing even more entities in the digital space under the new NIS 2 Directive.
Building cyber-resiliency has benefits for entire industries and countries. Preventing a member state or a competing organisation from suffering severe outages or data leakages benefits everyone on the continent. From our discussions with our clients — organisations understand the need for change and many are willing to adopt solutions if they exist. As part of this effort BranchKey is investigating methods of enabling intelligent cyber detection tools to be federated across organisations.
We take a strategy of the best data protection is when you do not connect it to the internet unless necessary precautions are taken.
Methods via Federated Data access allow clear dividers between systems preventing so called Horizontal side-channel attacks which make ransomware attacks so much more effective. Building redundancy into machine learning systems has yet to develop it’s own methods for defence — the industry is beginning to change.
If you’re interested in learning more about how federated learning is being applied in cyber resiliency check out some of these papers to get started:
- Detecting cyberattacks using anomaly detection in industrial control systems: A Federated Learning approach
- Anomaly Detection via Federated Learning
- Federated Learning for Cybersecurity: Concepts, Challenges and Future Directions
The problem is global and we’re working on an enabling system to allow organisations to collaborate with each other on these challenges. As this project develops there will be more to share along the way. If you’re looking for a solution as you build out your organisation’s security please reach out: [email protected] .
