Skip to content

Blog

Defense in Depth in Distributed Control Systems (DCS) - the network layer step by step

The network layer of defense in depth for a DCS: zones and conduits based on risk assessment, SL-T, firewalls between adjacent levels, the DMZ decision and microsegmentation.

Józef SulwińskiJózef Sulwiński

Published Updated 14 min read

Defense in DepthDCSOT securityICSnetwork segmentationIEC 62443SCADAzones and conduits
Defense in Depth in Distributed Control Systems (DCS) - the network layer step by step
Contents
  1. What the network layer protects in a DCS
  2. Step 1. List devices and data flows
  3. Step 2. Divide the system into zones based on risk
  4. Step 3. Set the SL-T for every zone and every conduit
  5. Step 4. Build the conduits and set the firewall rules
  6. Step 5. Decide whether a DMZ is needed
  7. Step 6. Monitor zone boundaries
  8. Step 7. Limit traffic within zones
  9. Step 8. Prepare to cut off a zone
  10. Checklist
  11. Where we help
  12. Technical basis
  13. Related articles
  14. Sources

In a DCS network where an office computer can connect to a controller, anyone who takes over that computer gets the same path. A single edge device with a vulnerability is then enough to send commands to the controllers. Defense in depth arranges security measures in successive layers, and the network layer decides how far an intruder gets.

The Dragos analysis, NIST guidance and IEC 62443 show that:

  • residents of Lviv had no heating for two days, and according to Dragos the router, management servers and heating controllers were not adequately segmented,
  • NIST recommends firewall rules that allow traffic only between adjacent network levels, in both directions,
  • under IEC 62443-3-2 the organisation groups assets into zones based on risk and sets a target security level for every zone and every conduit.

What the network layer protects in a DCS

A distributed control system (DCS) controls production at a single site, for example a refinery, a power plant, a wastewater treatment plant or a chemical plant. A supervisory controller sends set points to distributed controllers and collects data from them. The controllers drive actuators based on these commands and on sensor readings.

Fieldbus networks connect controllers to sensors and actuators without a separate wire to each device. According to NIST, dividing production into modules limits the effects of a single fault. The same description notes that a DCS is often connected to the corporate network so that business operations can see production data.

NIST describes defense in depth as five layers of security measures. The layers are meant to remove a single point of failure in the protection.

Layer according to NIST What the organisation does in it
1. Security management The organisation runs an OT security programme, assesses risk and sets the rules for the other layers.
2. Physical security The organisation protects control systems, tools and the surroundings against accidental or deliberate loss and damage.
3. Network security The organisation divides the network into zones, controls traffic at their boundaries and monitors communication.
4. Hardware security The organisation uses security features built into devices, for example a TPM, to protect device integrity and identity.
5. Software security The organisation uses application allowlists, installs patches and manages software configuration.

Layer 3 is built in eight steps. The other layers for SCADA and DCS are described in the article Defense in Depth for SCADA and DCS.

Step 1. List devices and data flows

List every device in the DCS network and every connection between them: source, destination, protocol, port and business purpose. Why: according to NIST, organizations typically build firewall rules from the map of data flows, and a rule without a recorded flow allows traffic that nobody needs.

Collect the flows passively, from a copy of traffic on a switch port (SPAN) or from a passive network tap. NIST names this listening mode as a possible first step of OT network monitoring. Why: passive capture sends no packets to the controllers, so it does not disturb the process.

The device inventory is described in the article on ICS asset inventory.

Step 2. Divide the system into zones based on risk

A zone is a group of devices that share common security requirements and have a clear boundary. A conduit is a controlled communication channel between zones. Under IEC 62443-3-2 the organisation groups assets based on an initial risk assessment or on other criteria, such as criticality, function, location, required access and responsible organisation.

Two partitioning rules are requirements in the standard:

  • control systems must be separated logically or physically from business systems, because they differ in function, owner and impact on health, safety and the environment,
  • safety systems must be placed in zones separated from the others, and where that is impossible, the whole shared zone must be treated as a safety zone.

The Purdue model, a hierarchy of levels from sensors to business systems, often serves as a starting point. IEC 62443-3-2 names it as an example basis for the division. In its example DCS architecture, NIST divides the network into four levels.

Level in the NIST example What it contains Purdue model levels
Field level This level holds sensors, actuators, PLCs and machine and process controllers. 0, 1 and 2
Operations management level This level holds the devices that monitor and manage the field level. 3
DMZ This level holds the servers that bridge the operations management level and the corporate network. The zone sits between level 3 and the corporate network.
Corporate network This level holds business systems and office computers. Levels above 3

NIST recommends considering separate segments for physical security systems, such as cameras, access control, card readers and VoIP telephony. The same example places IIoT (industrial internet of things) devices in a separate segment whose traffic passes through the firewall at the DMZ boundary.

Step 3. Set the SL-T for every zone and every conduit

SL-T (target security level) is a level from 1 to 4 that a zone or conduit is meant to achieve. Under IEC 62443-3-2 the organisation sets an SL-T for every zone and every conduit. The standard prescribes no method.

One of the methods described in the standard uses a risk matrix. The team starts from an estimated SL and assesses the risk with the countermeasures that this level implies. If the risk is unacceptable, raise the SL and repeat the assessment; the level at which the risk is acceptable becomes the SL-T.

Do not assign an SL by zone type, for example “a control zone is always SL 3”. Why: two control zones at different sites can have entirely different consequences of failure, and the SL-T follows from consequences and threats, not from the zone’s name.

A higher SL-T means more requirements for the network. IEC 62443-3-3 assigns successive system capabilities to successive levels.

SL Network segmentation Zone boundary protection
1 The system can logically separate the control network from other networks. The system monitors and controls communication at zone boundaries.
2 The system can also separate the control network physically. The boundary denies traffic by default and allows only permitted traffic.
3 The system provides network services to the control network without a connection to networks outside the control system. The boundary can cut off all communication (island mode) and blocks traffic when a boundary device fails.
4 The system can isolate critical networks logically and physically from the other control networks. The requirements are the same as for SL 3.

Step 4. Build the conduits and set the firewall rules

Place boundary devices, usually firewalls, between zones. Set the rules so that they deny traffic by default and allow only the flows recorded in step 1. Every rule has its source, destination, port and justification recorded.

Allow rules only between adjacent levels. NIST gives an example: a level 4 device does not communicate directly with level 2, 1 or 0 devices. Why: a compromised computer in the corporate network then cannot reach a controller without crossing further controlled boundaries.

Make outbound rules from the control network as strict as inbound rules. Why: according to NIST, control of outbound traffic varies widely in practice, and unmanaged outbound connections from lower levels are a significant risk.

Between the field level and the operations management level, NIST allows industrial-class firewalls. Such a firewall supports OT protocols and works in harsh conditions. A firewall that filters only addresses and ports will, once port 502 is open, pass every Modbus TCP command, including writes.

In Modbus TCP every request contains a function code that indicates whether the command reads or changes data. Write the rules at the field level boundary at function code level where the firewall supports it. The read and write codes are listed in the article on secure remote access to ICS.

When data only needs to flow upwards, for example from level 2 to a historian at level 3, consider a unidirectional gateway (data diode). Why: according to NIST, such a gateway between levels 2 and 3 may protect levels 0-2 from an incident at level 3 or above.

Step 5. Decide whether a DMZ is needed

A DMZ is a limited network of servers that joins two zones and controls the data flow between them. With it, the corporate network and the control network do not connect directly. IEC 62443-3-3 names a DMZ as an example of an effective boundary architecture, and the risk assessment decides whether to use one.

Build a DMZ when data passes between the corporate network and the control network, for example to a historian, a production planning system or a patch server. NIST then recommends that all communication between these networks pass through services in the DMZ. Why: the corporate server connects to a server in the DMZ, and only a server the organisation controls reaches the controllers.

Monitor and protect the services in the DMZ. Why: the DMZ connects to outside environments, and a compromised server in the DMZ lets an attacker move into the OT network without detection.

In the NIST example the operations management network has its own authentication server, separate from the one in the corporate network. Consider this design if the risk assessment justifies it. Why: compromising the account directory in the corporate network then does not automatically give accounts in the OT network.

Step 6. Monitor zone boundaries

Configure firewalls, switches and servers at zone boundaries to send event logs to a single, central location. Record the log retention period in a policy and provide storage for it.

Monitor traffic passively on boundary devices, from SPAN ports or taps. NIST notes that OT traffic is predictable, so deviations from the pattern are easier to detect than in an office network.

Place an intrusion prevention system (IPS), which blocks traffic on its own, at higher levels, for example on the DMZ interfaces. Why: NIST warns that an automated response in the control network can disrupt the process.

Observe encrypted traffic before encryption or after decryption. Why: an anomaly detection system cannot assess the content of traffic it cannot see.

Step 7. Limit traffic within zones

According to NIST, firewalls between zones do not limit traffic within a zone. A compromised engineering workstation can therefore connect freely to every controller in the same zone.

Microsegmentation divides a zone into smaller groups with their own rules. CISA describes it as an addition to defense in depth. In OT, use network-based microsegmentation, in switches and firewalls. Why: according to CISA, agent-based solutions may not be available for OT devices.

CISA proposes four phases, which are repeated for successive groups of devices:

  1. Select a group of devices to separate, based on criticality and ease of transition.
  2. Establish which systems the group must communicate with, and ask the people who operate it.
  3. Choose rules that allow the group to perform its function.
  4. Run the rules first in a mode that only flags violations, and only then in blocking mode. Why: the flagging mode reveals flows missing from the list before they stop the process.

The model in which every access is checked separately is described in the article on Zero Trust Architecture.

Step 8. Prepare to cut off a zone

During an incident it may be necessary to cut a zone off from the other networks. IEC 62443-3-3 recommends that after the cut-off the services needed to keep the installation running, or to shut it down in an orderly way, keep working.

Check which services from outside the zone the installation uses, for example address assignment (DHCP), name resolution (DNS) and a certificate authority. Duplicate these services in the control network. Why: a zone cut off from a DNS server in the corporate network can stop working at the very moment its isolation is most needed.

Rehearse the cut-off and measure how long it takes. The incident response plan records who makes the decision and who carries it out.

Checklist

  • All devices and flows in the DCS network are listed, and the flows were collected passively.
  • Control systems are separated from business systems.
  • Safety systems sit in separate zones.
  • Every zone and every conduit has an SL-T set in a risk assessment, and the result of the assessment is recorded.
  • Firewalls deny traffic by default and allow only traffic between adjacent levels.
  • Outbound rules from the control network are as strict as inbound rules.
  • Every firewall rule has its source, destination, port and justification recorded.
  • The DMZ decision follows from the risk assessment, and all communication with the corporate network passes through services in the DMZ.
  • Physical security systems and IIoT devices have separate segments.
  • Logs from boundary devices go to a central location, and traffic at the boundaries is monitored passively.
  • Microsegmentation within zones ran in violation-flagging mode before blocking was switched on.
  • The services needed after a zone cut-off run in the control network, and the cut-off has been rehearsed.

Where we help

During OT network segmentation we map the flows and design zones, conduits and firewall rules for a specific DCS. An IEC 62443 audit checks whether the SL-T follows from a risk assessment and whether zone boundaries meet the requirements for that level. OT network monitoring covers passive observation of traffic at zone boundaries, and an OT security assessment checks which paths lead from the corporate network to the controllers. You can reach us through the contact form.

Technical basis

Step / topic Where it is
Structure of a DCS NIST SP 800-82r3, 2.3.3
Five layers of defense in depth NIST SP 800-82r3, 5.1.2 and 5.2
1. Data flow map, passive monitoring NIST SP 800-82r3, 5.2.3.1 and 5.2.3.3
2. Definitions of zone and conduit IEC 62443-1-1, 3.2.116 and 3.2.27
2. Grouping based on risk IEC 62443-3-2:2020, ZCR 3.1
2. Separation of business and safety systems IEC 62443-3-2:2020, ZCR 3.2 and ZCR 3.3 (requirements)
2. Purdue model as an example basis for division IEC 62443-3-2:2020, 4.4.2.2; NIST SP 800-82r3, 5.2.3.1
2. Levels in the DCS example, physical security and IIoT segments NIST SP 800-82r3, 5.4.1 and 5.4.2
3. SL-T for every zone and conduit, no prescribed method IEC 62443-3-2:2020, ZCR 5.6
3. Network requirements for SL 1-4 IEC 62443-3-3, SR 5.1 and SR 5.2 with enhancements
4. Rules between adjacent levels, outbound rules, unidirectional gateway NIST SP 800-82r3, 5.2.3.1 and 5.4.1
4. Deny by default IEC 62443-3-3, SR 5.2 RE 1 (SL 2+)
4. Function code and port 502 MODBUS Application Protocol V1.1b3
5. Definition of a DMZ IEC 62443-3-3, 3.1.19
5. DMZ as an example boundary architecture IEC 62443-3-3, SR 5.2 (rationale)
5. Communication through DMZ services, separate OT authentication server NIST SP 800-82r3, 5.4.1
6. Logs, monitoring, IPS, encrypted traffic NIST SP 800-82r3, 5.2.3.2 and 5.2.3.3
7. Traffic within a zone NIST SP 800-82r3, 5.2.3.4
7. Microsegmentation phases, OT without agents CISA, Microsegmentation Part One, 3.1-3.4 and 4.7
8. Services after a zone cut-off, island mode IEC 62443-3-3, SR 5.1 (rationale) and SR 5.2 RE 2 (SL 3+)
Safety systems NIST SP 800-82r3, 5.3.1
  • Defense in Depth for SCADA and DCS - all protection layers, from physical security to third-party access
  • OT network segmentation - segmentation methods and the matrix of flows between zones
  • Zero Trust Architecture - checking every access and its place in the OT network
  • FrostyGoop - the course of the attack on the Lviv heating plant

Sources

  1. Dragos - FROSTYGOOP: ICS Malware Targeting Operational Technology - Dragos (accessed: )
  2. NIST SP 800-82 Rev. 3: Guide to Operational Technology (OT) Security (September 2023) - NIST (accessed: )
  3. CISA - Journey to Zero Trust: Microsegmentation in Zero Trust, Part One (29 July 2025) - CISA (accessed: )
  4. MODBUS Application Protocol Specification V1.1b3 - Modbus Organization (accessed: )

Need help in this area?

Our experts will help you assess the risk and plan next steps.

Talk to an expert

We'll discuss scope, methodology, and timeline.

Book a call