The short answer
Passive Discovery
Passive discovery copies the traffic on a control network to a sensor, through a switch mirror port or a network tap, and the sensor identifies devices, addresses, protocols, and conversations from what it sees, and often reads model and firmware information from the identification messages the protocols carry. It sends nothing, so it cannot disturb a controller the way an active scan can. It sees only traffic that crosses the point where it listens, so devices that are silent, on serial links, or on segments that do not pass the mirror are invisible to it, and it complements rather than replaces the site walk. Once the inventory exists, the same sensor watching the same traffic alerts on a new device, a new conversation, or a programming session that no change record explains.
Key points
- Active scans have crashed controllers; passive listening cannot, because it sends nothing.
- A mirror port or a tap at the core switch, a sensor with enough capacity, and time: that is the setup.
- The traffic reveals devices, addresses, protocols, who polls whom, and often model and firmware.
- Blind spots: silent devices, serial links, segments that do not cross the listening point.
- The sensor is itself an asset: hardened, its capture data protected, its management interface restricted.
- After discovery, the same sensor is continuous monitoring: alerts on anything new.
Why not scan
A network scanner finds devices by sending packets to every address and port and reading the responses. Office computers answer and carry on. Some controllers, remote units, and instruments were built before anyone imagined unsolicited traffic, and a burst of connection attempts to unusual ports can fault their network stack, drop their communication, or reboot them. The history of control system scanning includes stopped processes, and the caution is deserved. Passive discovery avoids the question: the sensor only receives, so nothing it does can reach a device.
The setup
- 1
Choose the listening point
The core switch through which most conversations pass: between the SCADA servers and the controllers, and between the plant and the remote site links. Several points may be needed on a segmented network.
- 2
Mirror or tap
A mirror session on a managed switch copies the traffic from the chosen ports or VLANs to the sensor port; a hardware tap sits inline on a link and copies it without depending on the switch. Mirrors are simpler; taps are more reliable and do not load the switch.
- 3
Size the sensor
Enough network capacity to receive the mirrored traffic without drops, enough storage for the capture, and analysis software: a purpose-built OT monitoring product, or an open-source capture and analysis stack for a utility with the skills.
- 4
Listen for a full cycle
A week at least, covering a backwash, a shift change, a weekend, a report run, and a vendor session, so that every periodic conversation is seen.
- 5
Analyze
The list of addresses and devices, the protocols each speaks, the conversation map of who talks to whom, the identification fields, and the anomalies: an address nobody recognizes, a protocol nobody expected, a conversation with the office.
- 6
Reconcile with the inventory
Every address matched to an inventory entry or investigated. The conversation map becomes the conduit list for segmentation.
What the traffic tells you
| Observation | What it reveals |
|---|---|
| Addresses and MAC addresses | Devices present; the vendor from the MAC prefix |
| Protocols per device | The role: a Modbus server is a device, a DNP3 master is the SCADA, a web session is a management interface |
| Identification messages | Model and firmware from device identification requests in Modbus, EtherNet/IP identity, SNMP system descriptions, and vendor protocol banners |
| Conversation map | Who polls whom, how often, with what functions; the basis of firewall rules and the baseline for anomaly detection |
| Programming and configuration sessions | Which workstations program which controllers, and when |
| Unexpected conversations | A controller talking to the office, a device reaching the internet, a laptop that appears on Tuesdays |
| Broadcasts and errors | Misconfigured devices, duplicate addresses, chatty protocols, loops |
| Cleartext credentials | Protocols carrying passwords in the clear, which is a finding in itself |
Blind spots
- Devices that never talk during the capture: a spare controller powered but idle, a backup radio, an instrument polled once a day.
- Serial links: RS-485 buses and serial radios carry no Ethernet frames; the gateway that fronts them appears, its devices do not.
- Segments that do not cross the listening point: a remote site switch, an isolated panel network, a vendor skid with its own switch.
- Traffic on the same switch that never reaches the mirror, depending on how the mirror session was configured.
- Encrypted protocols, which reveal endpoints and volume but not content.
- Everything the walk finds and the network does not: what is in the panel, what is powered off, what is not connected.
From discovery to monitoring
Once the inventory and the conversation map are established, the sensor keeps listening, and its job changes from discovery to detection. A new address, a new protocol from a known device, a conversation between two devices that never spoke, a programming session outside a change window, a controller mode change, a firmware download: each is an alert, matched against the change records like any detection. Purpose-built OT monitoring products do this with protocol knowledge and rules for the common controllers; a utility with a capture stack can build simpler alerts for new addresses and new conversations. Either way, the alerts go to a person who will look at them, or the sensor is decoration.
Frequently asked questions
- Will the mirror port affect the switch or the process?
- A mirror session adds load to the switch processor and the mirror port; on a modern managed switch mirroring a few ports of control traffic, it is negligible. Mirroring every port on a busy switch to one port drops packets at the mirror, which affects the sensor and not the process. A tap has no effect on the switch at all.
- How long should the capture run?
- A week for discovery, to catch weekly patterns; then indefinitely as monitoring. Storage for full packet capture fills quickly; monitoring products keep metadata and alerts rather than every packet, and a utility keeping raw capture retains a rolling window.
- Can passive discovery replace the site walk?
- No. It finds what talks; the walk finds what exists. A utility that does only the walk has an inventory without a conversation map; one that does only passive discovery has a map of the talkative devices and no idea what is in the panels. Do both, and reconcile.
- Is there a free option?
- Open-source packet capture and analysis tools, run on a spare machine at a mirror port, produce the address list, the protocol list, and the conversation map for a utility with someone able to read them. Purpose-built products add protocol decoding for control systems, device identification, and alerting, and are worth the cost when the sensor is to become monitoring.
Related topics
- Building an OT Asset InventoryThe list of everything in the control system with a network interface or a program, and how to build one that is right: what counts, the fields to record, the site walk, the switch tables and passive listening that find the rest, and keeping it current.
- Documenting Firmware VersionsWhy the firmware version of every controller, module, drive, switch, radio, and instrument belongs in the inventory with the date it was read: advisories are issued by version. Where to read each one, the several versions one controller has, and the cadence.
- Program Change DetectionKnowing when a controller program changed and whether it was supposed to: change counters and signatures the SCADA can alarm on, scheduled compares, audit logs, mode and network monitoring, what each catches, and what to do when one fires.
- Zones and ConduitsThe IEC 62443 way to segment a control system: grouping assets into zones with a shared security level, inventorying every conduit between them, and turning the drawing into firewall rules. With a worked water utility example.
- Firewall Rule DesignWriting the rule set for a control network boundary: default deny, one rule per conduit with its reason, hosts not subnets, protocol and port and function, direction and initiator, logging on every rule, and the review that removes forgotten rules.
- How to Configure VLANs on a Control NetworkTurn a zone plan into switch configuration: a VLAN per zone with its own subnet, access ports for devices, tagged trunks between switches, a native VLAN that carries nothing, a separate management VLAN, and routing between VLANs only through the firewall.
Direct contact
Have a controls question?
Reach Eric Sullivan directly about anything on this site, a controls or automation topic, or one of his personal projects.