Skip to main content

Zero-Trust Mesh Networking

EmberNet secures every connection between edge nodes, cloud services, and users with a zero-trust mesh. There is no site-to-site VPN to build, no port to forward, and no static firewall rule to maintain.

The principles​

Never trust, always verify. Every connection request is authenticated and authorized regardless of where it came from. There is no trusted network, and being inside the perimeter grants nothing.

Least privilege. Services and users get the minimum access they need. Policy is defined per service, per user, and per tenant, and the default is deny.

Assume breach. The architecture takes for granted that something will eventually be compromised. Identity is checked on every connection rather than only at the edge, so a foothold on one host does not become lateral movement across a flat network.

How it works for you​

Each node and service is issued its own identity when it joins. From there:

  • Nodes connect outbound to the fabric. Nothing dials in, so there is no inbound port to open and no public address on your equipment.
  • Every connection is mutually authenticated and encrypted, with credentials the platform issues and renews on its own.
  • Access follows identity, not network location. There is no address space to scan, which is what removes lateral movement.
  • Policy decides which identities may talk to which services.

The implementation is Flux, and Fireball manages it for you from within EmberCORE.

What this buys you​

PropertyWhat it means in practice
No inbound portsNo firewall exceptions, no port forwarding, no exposed equipment
No VPNNo site-to-site tunnels to build, document, or maintain
Automatic certificatesIssued when a node joins, renewed before expiry, revocable instantly
NAT traversalWorks through carrier-grade NAT with no configuration
Network agnosticRuns over cellular, satellite, WiFi, or ethernet
Everything over HTTPSPasses through restrictive client networks without special handling

Access policies​

A policy grants one identity access to a service, and everything not granted is denied. Fireball sets policy for you, in four places:

TypeControls
Service to serviceWhich platform services can reach each other
User to serviceWhich users can reach which services
Tenant to tenantCross-tenant traffic, denied by default
External accessAccess originating outside the mesh

Certificates​

Certificates are issued automatically when a node joins, renewed before they expire, and revoked immediately if a host is ever compromised. Because they are short lived, a stolen certificate is of little use to anyone.

Accurate time matters for that reason. Keep NTP working on every node, because a clock that has drifted looks like a network problem rather than a time problem.

Monitoring and troubleshooting​

EmberCORE shows your organization's Flux services, identities, and edge routers, and measures the latency to each service, so a path that has gone slow or dark shows up before somebody calls about it.

SymptomLikely causeWhat to do
Node won't join the meshOutbound HTTPS blockedVerify outbound HTTPS from the node itself
High latencyCongestion or routingCheck the network path and available bandwidth
Certificate errorsClock skewFix NTP on the node
Access deniedNo policy grants itReview policy; remember the default is deny

Access denied is the one that trips people up, because it looks like a failure and is usually the system working correctly. Check for a missing grant before suspecting anything else.

Next steps​