Skip to main content

Plasma edge lockdown

Plasma's edge firewall puts the box's traffic under one policy: what comes in from the plant network, what goes out to it, and what may reach the PLCs and I/O behind the box. It is a container on the box, and it updates itself from the stable channel like the rest of the endpoint. See EmberRTOS for how it is switched on.

The default policy on the plant network​

DirectionAllowed
In from the plant networkReplies to connections the box started, and nothing new. No service listens on the plant network.
Out to the plant networkTCP 443, DNS, DHCP, NTP, and ping

Everything else is dropped and counted. The box's services are reached over Flux or on the PLC leg, never from the plant LAN. Port 443 is the one path off the site, which is the path Flux already uses.

Protecting the PLCs​

When a box has a PLC leg, Plasma sits in line between the box and the plant switch, and between the box and the PLC switch. Nothing on a PLC needs reconfiguring, and industrial protocols that depend on broadcast keep working. A box whose one port carries the PLC networks as VLANs is guarded too, one VLAN at a time, even where several VLANs reuse the same addresses.

Each PLC has its own entry in the box's policy: its name, address, and the protocols it speaks, such as Modbus, EtherNet/IP, Profinet, OPC UA, or CODESYS. Plasma matches a PLC by its address and its hardware address together, so a device plugged into the PLC leg cannot reach a controller by copying an IP address. A device or protocol that is not listed is dropped and counted, and the plant network and the PLC leg are never connected to each other through the box.

If your PLCs share the plant network with no separate leg, Plasma protects the box and what the box exposes, but it cannot filter traffic between devices that never cross the box.

Allowing more: the tenant allowlist​

A tenant that needs its boxes to reach something beyond the defaults, such as a site-specific server, adds it to the tenant's allowlist in Plasma. An entry names a destination, by address or by name, and a port. The allowlist applies to every box in that tenant and never affects another tenant. Anything not on the list stays dropped.

A new policy is checked before it is applied, applied in one step, and the last good policy is kept. A policy that does not validate never reaches the wire.

A restart costs the plant nothing​

A planned Plasma restart, such as an automatic update, keeps the box's path to the plant in place and keeps its network address. The traffic that was flowing keeps flowing, so a historian sees no gap.

If Plasma stops​

A stopped control loop stops a machine, so the box fails in the safe direction:

  • Management and Flux traffic, and the cyclic traffic between the box and its PLCs, keep flowing under a fallback policy that allows the same traffic and nothing more.
  • Anything that would pass between the plant network and the PLC leg stays closed.
  • Losing the link to Plasma's controller changes nothing on the wire. The box keeps enforcing the policy it has.

The box restores its normal path on its own, without a person.

Seeing what was refused​

Every drop is counted. The counts reach EmberCORE through EmberSecurity every 30 seconds and appear in the Blocked traffic and Threat map views of the EmberSecurity area, read only. A count that suddenly rises is usually a device that is new, misconfigured, or not supposed to be there.

Next steps​