A loop has formed in the ring, a broadcast storm has started and you want to connect to the switch via SSH. You can’t, because your management traffic passes through the very network that is the fault. Switches, firewalls, routers and protocol gateways in the OT network carry two types of traffic: production traffic and the traffic used to manage the devices themselves. In most plants, both flow over the same network, even the same VLAN. Below, we describe separating the management plane of OT network devices from production: management VLAN and VRF, out-of-band network and console server, hardening management protocols and the location of the management station. Segmentation of production traffic is covered in other articles; the subject here is management access only.
Why the management plane should be separated
There are three concrete reasons.
Attack surface: If every device in the production VLAN can reach a switch’s SSH, web interface or SNMP service, a compromised HMI or engineering workstation can target network devices. Whoever takes over a network device can redirect traffic, eavesdrop via port mirroring or disrupt the ring structure. One device, the whole line.
Access during a fault: When there is a loop, broadcast storm or faulty VLAN change on the production network, it may not be possible to reach the device over the same network. The access needed to fix the fault must not be affected by the fault.
Change safety: When VLAN and address changes are made in segmentation projects, the management connection must stay outside the structure being changed. If it doesn’t, a wrong command cuts off access to the device and the job comes down to going to site. This is a common scenario: the management VLAN is accidentally removed from a trunk, the session freezes, and the device is in a cabinet far away. That is why the management network is the first wave in the plan for segmenting a flat OT network.
The in-band management VLAN and VRF approach
The most common first step is to move the management interfaces of all network devices into a separate management VLAN. On access switches, the management IP address is defined only on this VLAN’s interface; the switch has no IP address in the production VLANs. Because the management VLAN is carried over trunks, it physically uses the same links; that is why it is called “in-band”.
On a Layer 3 switch or router, however, the management VLAN’s interface is by default in the same routing table as the production VLANs. Unless ACLs are written, the management network can be reached by routing from production VLANs. A separate VLAN does not mean a separate network. VRF (Virtual Routing and Forwarding) solves this by creating a separate routing table for management: there is no path between interfaces in the management VRF and production interfaces unless route leaking is explicitly defined.
Not every device supports VRF. Many Layer 2 industrial switches offer only a single management interface and default gateway; on these devices, separation is achieved with the management VLAN and VRF or ACLs on the Layer 3 device routing it. Some devices have a physically separate management port; this port usually works with its own routing table and is suitable for connecting to an out-of-band network.
The out-of-band network and console server
In out-of-band (OOB) management, management traffic uses none of the links or devices of the production network. It has two components:
- Separate management network: The devices’ separate management ports are connected to management workstations via switches installed for this purpose only. This network is not bridged to the production network at any point.
- Console server: The serial console ports of critical devices are connected to a console server. Even if the device’s IP configuration is completely broken, a software update is interrupted or the device hangs at boot, it can be accessed from the console. An open console session during a firmware update is one way of seeing why a device won’t boot without going to site.
OOB has a cost: extra cabling, extra switches, console servers and their maintenance. That is why it is applied gradually. Core switches, IT/OT boundary firewalls, cell firewalls and switches in the ring manager role are put on OOB; the many access switches inside cells are managed via the in-band management VLAN.
At remote sites, adding backup access to the console server via a cellular modem is a common choice. This adds a new external connection point to OT; it is kept off by default, connects only to the organisation’s management point via VPN or a private APN, and its use is logged.
Hardening management protocols
Even if the management network is separated, the management services on the device are hardened separately. A separate network does not turn off an open Telnet.
- SSH: SSH version 2 only; Telnet disabled. Access is protected with an ACL restricted to the management subnet.
- Web interface: Disabled if not used; if used, HTTPS only and the management subnet only.
- SNMP: SNMPv1/v2c community names are removed; SNMPv3 authPriv is used. The SNMPv3 user-based security model is defined in RFC 3414. Write access is not granted unless needed. Industrial switches still running with factory-default community names are found in the field; this is one of the first things to check in an inventory scan.
- Authentication: Central authentication with TACACS+ or RADIUS; a single local emergency account to be used when the server cannot be reached.
- Unused services: Whether CDP/LLDP is needed on production ports is assessed. Because PROFINET topology diagnostics rely on LLDP, disabling LLDP in PROFINET networks breaks diagnostic capability.
- Time and logs: NTP and syslog targets are on the management network; log timestamps are consistent.
For details of SNMPv3 configuration, the SNMPv3 configuration guide can be used.
Location of the management station and tools
The management network is only as secure as the station that accesses it. Stations and tools that manage network devices are placed according to these principles:
- Inside OT, in a separate zone: The management station sits in OT, in a zone with access only to the management network. Management is not done directly from corporate laptops.
- Remote access via the DMZ: If remote management is needed, access is via the jump host in the DMZ, opening a session on the management station. This placement is based on the principles in the article OT DMZ design.
- Monitoring and backup tools: The SNMP monitoring server, syslog collector and configuration backup tool also sit on the management network or in the OT server zone that can reach it only on the necessary ports.
- Laptops: Maintenance laptops connected with a console cable in the field should be OT-dedicated devices that do not connect to the corporate network. A laptop that reads email in the office during the week and connects to the core switch’s console at the weekend is a bridge carried between two networks.
When choosing an industrial switch, management features (separate management port, VRF, SNMPv3, TACACS+/RADIUS support, console port) are checked in advance; for the selection criteria, see the article industrial switch selection.
Comparison table and configuration example
| Criterion | Flat (no separation) | In-band VLAN/VRF | Out-of-band |
|---|---|---|---|
| Access from production | On | Closed with ACL/VRF | Physically absent |
| Access during a fault | Affected by the fault | Partly affected | Unaffected |
| Additional hardware | None | None or little | Switches, console server |
| Where it fits | Not recommended | Access switches | Core, firewall, ring manager |
The example below shows a management VRF, SSH restriction and SNMPv3 configuration on a Cisco IOS XE-based Layer 3 switch. Addresses and names are examples; commands may differ depending on platform and software version.
hostname OT-CORE-1
ip domain name otnet.example.com.tr
!
vrf definition MGMT
address-family ipv4
exit-address-family
!
interface Vlan900
description OT-MGMT
vrf forwarding MGMT
ip address 10.90.0.11 255.255.255.0
!
ip route vrf MGMT 0.0.0.0 0.0.0.0 10.90.0.1
!
crypto key generate rsa modulus 2048
ip ssh version 2
ip access-list standard MGMT-STATIONS
permit 10.90.10.0 0.0.0.255
!
line vty 0 15
access-class MGMT-STATIONS in vrf-also
transport input ssh
login local
!
no ip http server
ip http secure-server
!
snmp-server group OTMON v3 priv
snmp-server user nms01 OTMON v3 auth sha <auth-password> priv aes 128 <priv-password>
snmp-server host 10.90.10.20 vrf MGMT version 3 priv nms01
logging host 10.90.10.30 vrf MGMT
ntp server vrf MGMT 10.90.10.5
After configuration, three tests are carried out: does SSH access from the management station work, is an SSH attempt from a device in the production VLAN rejected, and can the SNMP monitoring server query the device with SNMPv3? Doing only the first test and leaving is a common shortcoming; the real proof is the rejected attempt. Check that the output of show ip route vrf MGMT contains only management networks. In production, a TACACS+ or RADIUS-based AAA configuration should be used instead of the login local line in the example.
For continuous monitoring of the management network and noticing drift in network devices, see our industrial network monitoring and fault analysis page, and for addressing the management plane together with segmentation design, our OT cybersecurity and segmentation page.
Frequently asked questions
Is out-of-band management necessary in a small OT network?
Not for every device. In small networks, a separate management VLAN and rules restricting management access to specific workstations meet most needs. Out-of-band access is considered at least for devices that cannot be recovered without going to site if access is lost, such as core switches, boundary firewalls and the ring manager.
What is the difference between a management VLAN and a VRF?
A management VLAN puts management interfaces into a separate broadcast domain at Layer 2; however, on a Layer 3 switch, this VLAN’s interface shares the same routing table as the other VLANs. A VRF creates a separate routing table for management; unless explicitly defined, the management network cannot be reached by routing from production VLANs. VRF support depends on the device model and software licence.
Is it safe to access the console server via a cellular modem?
It is acceptable if designed correctly, but it adds a new external connection to OT. The modem should not present a management interface open to the public internet; it should connect only to the organisation’s management point via a private APN or VPN, be off by default and have its use logged. This connection is added to the OT security risk assessment as a separate conduit.
Why use SNMPv3 instead of SNMPv2c?
SNMPv2c sends the community name in clear text, and authentication relies on this name alone. Someone listening on the network can learn the community name; if a community name with write access is captured, the device configuration can be changed. SNMPv3 in authPriv mode provides user-based authentication and encryption.