OT Asset Inventory with Passive Network Monitoring
Build an OT asset inventory safely: SPAN and TAP placement, Zeek with CISA ICSNPP parsers, Wireshark filters for ICS protocols and which attributes to record.
Every OT security framework starts in the same place: know what you have. IEC 62443 needs an inventory to draw zones, NIST SP 800-82 lists Asset Management first under Identify, and you cannot do vulnerability management or incident response on devices you do not know exist.
In IT you would run an active scanner and an agent. In OT that is often a bad idea: some controllers and network modules handle unexpected traffic poorly, and a scan during production can cause faults. This guide shows how to build an inventory mostly from passive network monitoring, then fill the gaps safely.
Why passive first
Passive discovery only observes existing traffic. It adds no packets to the control network, so it cannot crash a device. ICS protocols also help: they are chatty and usually unencrypted, so passive capture reveals much more than in IT. A single EtherNet/IP ListIdentity response contains vendor, device type, product code, firmware revision, serial number and product name. Modbus Read Device Identification (function code 43, MEI type 14) returns vendor and product strings. S7comm SZL reads expose module order numbers and firmware.
Passive has limits: devices that do not talk during your capture window are invisible, serial-only devices behind gateways do not appear at all, and you see network identity rather than physical location. That is why the inventory combines passive data with documents and walkdowns.
Step 1: define scope and attributes
CISA and partner agencies published Foundations for OT Cybersecurity: Asset Inventory Guidance for Owners and Operators in August 2025. It recommends a five-step process: define scope and objectives, identify assets and collect attributes, create a taxonomy, manage and collect data, and implement life cycle management. It also lists high-priority attributes to record for each asset:
| Attribute | Passive network source |
|---|---|
| IP address, MAC address | Any traffic, ARP |
| Hostname | DHCP, DNS, NetBIOS, protocol identity fields |
| Manufacturer, model | MAC OUI, EtherNet/IP identity, Modbus device ID, S7 SZL |
| Operating system / firmware | Protocol identity, banners, SMB, HTTP headers |
| Active communication protocols | Observed flows and parsers |
| Ports / services | Observed flows |
| Asset role / type | Inferred from protocol role (client or server, master or outstation) |
| User accounts | Partially, from authentication traffic |
| Logging | Not visible passively |
| Asset criticality, asset number, physical location | Not visible passively: from engineering and maintenance records |
Agree on this attribute list before you collect anything, so every source fills the same fields.
Step 2: choose capture points
Where you capture matters more than which tool you use. Priorities:
- Supervisory segments (Purdue Level 2): HMI, SCADA servers and engineering workstations talking to controllers. This is where most identity information flows.
- Control segments (Level 1): controller to controller and controller to I/O traffic, often multicast or implicit I/O.
- Zone boundaries and the IDMZ: shows every conduit, which you need for network segmentation.
- Remote access paths: jump hosts and vendor connections.
Two ways to get the traffic:
- SPAN (port mirroring) on a managed switch. Cheap and fast to deploy, but it can drop packets under load, may not copy malformed frames, and changing switch configuration in production needs a change request.
- Network TAP inline on a link. A passive or fail-open TAP copies everything with no switch involvement, but installing it means briefly breaking the link, which must be scheduled during an outage window.
Make sure the monitoring port is receive-only on the sensor (no IP address, no transmit) so the sensor cannot be used as a path into the control network. On Linux:
# Bring up the capture interface without an IP address
sudo ip link set dev enp3s0 up promisc on
sudo ip addr flush dev enp3s0
# Disable offloads that distort captures
sudo ethtool -K enp3s0 gro off lro off tso off gso off
Step 3: capture and parse with Zeek and ICSNPP
Zeek ships parsers for Modbus and DNP3 out of the box and writes them to modbus.log and dnp3.log alongside conn.log. CISA's ICSNPP project (Industrial Control Systems Network Protocol Parsers) adds open source Zeek parsers for many more ICS protocols, including EtherNet/IP and CIP, S7comm and S7comm-plus, BACnet, OPC UA Binary, PROFINET IO CM, HART-IP, Omron FINS and others, plus extended logging for Modbus and DNP3.
Install the packages you need with the Zeek package manager:
zkg refresh
zkg install icsnpp-modbus icsnpp-enip icsnpp-s7comm icsnpp-bacnet icsnpp-opcua-binary
# Confirm the S7comm plugin is loaded
zeek -N | grep -i s7comm
Run Zeek offline against a capture first, which is the safest way to start:
# Record 24 hours in hourly files on the sensor
sudo tcpdump -i enp3s0 -n -s 0 -G 3600 -w 'ot-%Y%m%d-%H%M.pcap'
# Parse a capture with installed packages loaded
zeek -C -r ot-20261110-0800.pcap local
ls *.log
# conn.log modbus.log modbus_detailed.log enip.log cip.log cip_identity.log
# s7comm.log cotp.log dnp3.log ...
Now extract inventory data from the logs with zeek-cut:
# Every host that spoke Modbus, and in which role
zeek-cut id.orig_h id.resp_h id.resp_p < modbus.log | sort -u
# EtherNet/IP identity: vendor, device type, firmware, serial, product name
zeek-cut socket_address vendor_name device_type_name product_name \
revision serial_number < cip_identity.log | sort -u
# Talkers per ICS port (502 Modbus, 102 S7/MMS, 44818 and 2222 EtherNet/IP,
# 20000 DNP3, 2404 IEC 60870-5-104, 4840 OPC UA, 47808 BACnet)
zeek-cut id.orig_h id.resp_h id.resp_p proto < conn.log \
| awk '$3 ~ /^(502|102|44818|2222|20000|2404|4840|47808)$/' \
| sort | uniq -c | sort -rn | head -50
Look at the field headers in each ICSNPP log before scripting against it, since field sets differ between protocols and package versions.
For a packaged deployment, CISA's open source Malcolm tool bundles Zeek, the ICSNPP parsers, Suricata and a web interface for browsing the results.
Step 4: targeted inspection with Wireshark
Wireshark is ideal for confirming what Zeek reports and pulling identity strings by hand. Useful display filters, using field names from the Wireshark display filter reference:
# Modbus/TCP traffic and writes (FC 5, 6, 15, 16)
mbtcp
modbus.func_code in {5 6 15 16}
# Modbus Read Device Identification (FC 43 / MEI 14)
modbus.func_code == 43
# EtherNet/IP ListIdentity and the identity fields in replies
enip.command == 0x0063
enip.lir.vendor || enip.lir.prodcode || enip.lir.name
# S7comm jobs and SZL reads that return module identity
s7comm.header.rosctr == 1
s7comm.data.userdata.szl_id
# DNP3 application function codes
dnp3.al.func
Use Statistics > Conversations and Statistics > Endpoints for a quick per-host view, and Statistics > Protocol Hierarchy to spot unexpected protocols (SMB, RDP, HTTP or DNS where you did not expect them).
Step 5: fill the gaps carefully
Passive data will not show everything. Fill gaps in this order, from safest to riskiest:
- Documents: control system drawings, I/O lists, network diagrams, PLC project files, maintenance records and purchase records.
- Physical walkdowns of cabinets and control rooms: model, serial, firmware label, location and network connections. This is the only way to find serial devices and unmanaged switches.
- Native tools and configuration exports: switch MAC address tables and ARP tables, DCS system diagnostics, engineering software device lists.
- Targeted active queries using the device's own protocol, at low rate, from an approved host, during a maintenance window, and after testing on a spare or lab device. NIST SP 800-82 Appendix E describes combined passive and active approaches for this reason.
Avoid broad TCP scans of Level 1 and Level 0 networks unless the asset owner and the control vendor have approved them.
Step 6: build the taxonomy and keep it current
An inventory becomes useful when you add structure:
- Criticality: what happens to the process if this asset is lost or manipulated.
- Function: controller, HMI, historian, engineering workstation, network device, safety controller.
- Zone and conduit: which zone it belongs to and which conduits it uses, following 62443 zones and conduits.
- Relationships: which HMI talks to which controller, which engineering station programs which PLC. These communication pairs become your firewall allow lists.
Store the result in one place (a CMDB with OT fields, a dedicated OT asset tool, or a controlled spreadsheet as a start) and tie updates to change management: every replacement, firmware update or new device updates the record, including emergency changes.
Keep the sensor running after the project. Continuous monitoring turns the inventory into a detection source: a new MAC address on a control segment, a new Modbus client issuing writes, or a program download outside a maintenance window are all events worth alerting on. These map to techniques in MITRE ATT&CK for ICS such as Remote System Discovery (T0846), Network Sniffing (T0842) and Program Download (T0843).
Checklist
- Attribute list agreed with operations and engineering.
- Capture points chosen at Level 2, Level 1, zone boundaries and remote access paths.
- Receive-only sensors with no IP on the capture interface.
- Zeek with ICSNPP parsers (or Malcolm) parsing at least 24 hours to cover shift and batch cycles.
- Results cross-checked with Wireshark, documents and walkdowns.
- Gaps filled with approved, low-impact methods only.
- Inventory stored centrally, linked to change management, and reviewed on a schedule.
References
- CISA: Foundations for OT Cybersecurity, Asset Inventory Guidance for Owners and Operators
- CISA ICSNPP: Industrial Control Systems Network Protocol Parsers
- ICSNPP Modbus extension for Zeek
- ICSNPP S7comm parser for Zeek
- ICSNPP EtherNet/IP and CIP parser for Zeek
- CISA Malcolm network traffic analysis suite
- Zeek package manager quickstart
- Zeek Modbus analyzer script reference
- Wireshark display filter reference: Modbus
- Wireshark display filter reference: EtherNet/IP
- IANA Service Name and Transport Protocol Port Number Registry
- NIST SP 800-82 Rev. 3, Guide to OT Security