How invisible networks and zero trust cybersecurity protect shared science

The border problem

What happens when one experiment links a normal cloud server, an edge sensor, and an experimental quantum computer across different organisations? In the past, computer security worked like a medieval castle wall: once a user passed the outer perimeter, systems trusted them. But modern science shares data across institutions, making outer firewalls useless.

Hiding in plain sight

To solve this, researchers designed a framework applying zero trust cybersecurity, where no user or device is trusted by default. Instead of building walls, they turned to a secret speakeasy approach. By using darknet-inspired principles, computing nodes stay invisible to scans on the public internet, meaning attackers cannot target doors they cannot see.

The system is deployment-agnostic, running smoothly whether it is installed on high-performance supercomputers or small edge nodes. Four open-source tools drive the setup. An overlay network called Nebula links the machines invisibly. Keycloak confirms the user's identity, Open Policy Agent decides what they are allowed to see, and Traefik directs traffic securely to the right resource.

Testing the pieces

The team validated the design with a pilot deployment. Through policy tests and performance benchmarks, they measured how cleanly the framework handled cross-organisation access. They showed that this minimal software stack reduces operational complexity while keeping data flexible and tightly shielded across very different computer environments.

What we still do not know

This design was tested as a pilot rather than a full-scale production system run across dozens of global facilities. How it handles massive real-world traffic under heavy workloads remains unknown. Additionally, while the team plans to add post-quantum cryptography to withstand future quantum computer attacks, the real performance cost of those future encryption tools has not yet been measured.