An IT vulnerability scan that is routine on a server estate can freeze a PLC, trip a drive or drop an HMI off the network, so active testing in OT needs its own rules of engagement. Many controllers, protection relays and embedded devices have small network stacks, limited CPU and no tolerance for malformed or high-rate traffic, and their failure has physical consequences. This guide sets out when active scanning is justified, how to scope, approve and tune it, which techniques are lower and higher risk, how to stop and roll back, and how to measure that testing finds issues without causing them. It is written for OT security engineers, asset owners, control engineers and the internal or external testers they authorise.
Key takeaways
- Passive first, active by exception. Use passive monitoring, configuration data and vendor tools to build the inventory, then use targeted active queries only to fill specific gaps.
- Operations own the stop button. No active test runs without a named operations authoriser, a maintenance window or a process state where loss of the device is acceptable, and an agreed abort signal.
- Test on a replica before production. Prove every scan profile against the same model and firmware in a lab or test cell before it touches a live controller.
- Throttle and narrow. One probe at a time, small target sets, known ports only, and no service fingerprinting or brute forcing against Level 0 to 1 devices unless proven safe.
- Measure harm as well as findings. Track device faults, communication alarms and operator complaints during every test, and feed them back into the profile library.
Why OT devices fail under scanning
NIST SP 800-82 Rev. 3, the Guide to Operational Technology Security, distinguishes passive and active scanning in its network monitoring appendix and repeatedly advises testing changes on a test system before they reach the OT system. The reasons are practical:
- Limited network stacks: older controllers and communication modules can stop responding, reboot or drop into fault when they receive malformed packets, many simultaneous connections or unexpected protocols.
- Real-time scan cycles: controller CPUs prioritise the logic scan; extra communications load can lengthen scan times or trigger watchdog faults.
- Session limits: many devices support only a few concurrent connections, so a scanner can lock out the HMI or engineering station.
- Write-capable protocols: Modbus, DNP3, S7comm and similar protocols carry read and write functions on the same port without authentication. A careless tool that issues writes or mode changes is not scanning, it is control.
- Serial and gateway links: traffic forwarded through protocol gateways to serial devices can saturate low-bandwidth links.
Discovery activity is also what adversaries do. ATT&CK for ICS lists Remote System Discovery (T0846), Port Scan (T0846.001) and Remote System Information Discovery (T0888), and the same traffic can cause Denial of Service (T0814) or Device Restart/Shutdown (T0816) as side effects. Authorised testing must therefore be announced to the SOC so it is not mistaken for an attack, and so a real attack is not mistaken for the test.
Choosing the right method
Most OT inventory and vulnerability questions can be answered without sending traffic to devices. The table compares methods by risk and what they reveal. See the OT Asset Visibility Guide for the inventory programme this supports.
| Method | Risk to process | What it reveals | Use when |
|---|---|---|---|
| Passive network monitoring (SPAN or TAP) | None to the devices when the tap is installed safely | Talkers, protocols, firmware and model strings visible in traffic | Always first; continuous |
| Configuration and document review | None | Engineering project data, IP plans, firmware versions in project files | Alongside passive monitoring |
| Vendor engineering or asset management tools | Low when used as designed | Model, firmware, module inventory through the vendor's own protocol | Gaps in passive data for one vendor |
| Selective native protocol queries | Low to medium | Identity data requested with read-only native commands | Targeted gaps, after lab proof |
| Host-based assessment on Windows and Linux OT hosts | Low to medium | Patches, software, local accounts | Level 2 and 3 hosts with vendor approval |
| Network port and service scanning | Medium to high on Level 0 to 1 | Open ports and services | Level 3 and DMZ; lower levels only in windows |
| Authenticated vulnerability scanning | Medium on hosts, high on embedded devices | Missing patches and misconfigurations | Level 2 to 3 hosts, tested profile |
| Penetration testing and fuzzing | High | Exploitable paths, robustness | Lab or test cell; production only with explicit acceptance |
- Small sites: a passive sensor for a few weeks plus project file review usually produces a complete enough inventory; active scanning is rarely needed below Level 2.
- Large estates: build a profile library per device family once in a lab, then reuse it across sites under change control.
- Regulated or safety environments: safety instrumented systems and protection relays are normally excluded from active testing in production. See the Safety Instrumented Systems Security Guide.
Rules of engagement
NIST SP 800-115, the Technical Guide to Information Security Testing and Assessment, describes planning, rules of engagement and coordination for technical testing. In OT the rules of engagement must add process safety. A usable rules of engagement document covers:
- Purpose and question: exactly what the test must answer (for example, firmware versions on a PLC family that passive monitoring cannot see).
- Scope: explicit IP addresses or device tags, zones and conduits; anything not listed is out of scope. Exclude safety systems, protection relays and devices on serial links unless separately approved.
- Approvals: the asset owner, operations shift lead, control engineering, and the security lead sign. External testers also sign the contract terms.
- Timing: maintenance window, shutdown, or a process state agreed as tolerant; never during start-up, batch transitions or peak demand.
- Permitted techniques: named tools, versions and profiles from the tested library; banned techniques listed (no writes, no mode changes, no credential guessing, no denial of service tests).
- Communication: SOC and control room notified; a single bridge or radio channel; a defined abort word.
- Stop criteria: any device fault, communications alarm, unexpected HMI behaviour, or operator concern stops the test immediately.
- Recovery plan: spares, backups of affected controllers and an engineer on site able to restore. See the OT Backup, Recovery and Golden Image Guide.
- Data handling: results, packet captures and credentials stay inside the OT environment or an approved secure store, and are deleted by the tester after hand-over.
Tuning active scans for fragile devices
Where active scanning is justified, the aim is the lowest possible load and the narrowest possible set of interactions. The settings below are expressed generically; map them to whichever approved scanner you use and prove them in the lab.
| Setting | Safer choice | Why |
|---|---|---|
| Parallelism | One probe at a time per target; few targets at once | Avoids exhausting connection tables and CPU |
| Delay between probes | Hundreds of milliseconds or more | Keeps load below what the device stack can handle |
| Port list | Known ports for that device family only | Full port sweeps produce many unexpected connections |
| Service and version detection | Off for Level 0 to 1, or replaced by native read-only identity queries | Fingerprinting sends unusual payloads that older stacks mishandle |
| Scripts and plugins | Only reviewed read-only checks; no brute force, no denial of service or "unsafe" categories | Plugins may write, authenticate repeatedly or crash services |
| Host discovery | Use the known address list, not sweeps of whole subnets | Sweeps hit unknown devices with no tested profile |
| Credentials for authenticated scans | Dedicated read-only account, enabled only for the window | Limits damage if the scanner or its credential is compromised |
The Nmap reference guide is a useful illustration of how timing affects load: its "polite" template probes serially with a 400 millisecond delay, and its author notes that leaving out version detection reduces crash and bandwidth problems more than slowing timing does. Those observations come from general networks; OT devices need their own lab proof.
Scanner credentials and API keys are non-human identities with privileged reach across the plant. Store them in a vault, issue them per engagement, disable them afterwards, and log their use, in line with NIST SP 800-53 RA-5(5) for privileged scanning access. See the OT and ICS Identity and Access Guide.
Phased rollout: from lab to production
- Build the lab: spare controllers and modules matching production models and firmware, a representative HMI, and a switch with a capture port.
- Define profiles: one scan or query profile per device family, recorded with tool version, settings and expected traffic.
- Prove in the lab: run each profile repeatedly while monitoring CPU load, scan time, communications status and fault logs on the device. Any fault disqualifies the profile.
- Pilot on a non-critical unit: a redundant line, standby unit or test cell, with operations present.
- Expand in rings: extend to wider scope one zone at a time, reviewing results and harm indicators after each ring.
- Roll back: if a device faults, stop the run, return the device to a known state from its backup if required, record the event, and remove or revise the profile before any retry.
- Institutionalise: put the profile library under change control and re-prove profiles after firmware upgrades.
Penetration testing in OT
Penetration testing (NIST SP 800-53 CA-8) in OT has the same goals as in IT but different limits. Exploitation and robustness testing belong in a lab or factory acceptance environment; production testing should concentrate on the paths into OT, such as remote access, jump hosts, the IT/OT DMZ and engineering workstations. CISA's OT advisories repeatedly show these as the routes attackers use, for example limiting ICS connections to approved engineering workstations and protecting remote access with MFA. See the OT Penetration Testing Guide for engagement design.
- Prefer white box: give testers drawings and asset lists so they do not need to discover the network by force.
- Test the conduits: verify firewall and data diode rules from both sides rather than probing every device behind them.
- Use independent testers carefully: CA-8(1) independence is valuable, but external testers must also be accompanied by site engineers when on OT networks.
- Protect the findings: a report listing exploitable controllers is a sensitive document; restrict its distribution.
Pre-test and post-test checks
The difference between a safe test and an incident is often the ten minutes before it starts. Use a fixed checklist for every window, signed by the tester and the operations authoriser.
| When | Check | Owner | Evidence |
|---|---|---|---|
| Day before | Scope list matches the current asset inventory; no new devices in range | OT security | Signed scope sheet |
| Day before | Current backups exist for every in-scope controller and HMI | Control engineering | Backup log with hashes |
| Start of window | Process in the agreed state; redundant units healthy; no active alarms | Shift lead | Control room log entry |
| Start of window | SOC and monitoring team told the source address and time window | Tester | Ticket reference |
| Start of window | Baseline captured: controller mode, program checksum, CPU load, communications status | Control engineering | Screenshots or diagnostic export |
| End of window | Same baseline re-captured and compared; no change in mode or logic | Control engineering | Comparison record |
| End of window | Scanner credentials disabled; temporary firewall rules removed | OT security | Change closure |
Temporary firewall openings for the scanner are a frequent leftover. Raise them as a time-boxed change with an automatic expiry where the firewall supports it, and confirm their removal as part of closure. The same discipline applies to any jump host or laptop brought in for testing: it should be scanned and approved before connection and removed afterwards.
Detection and measurement
- During the test: watch controller diagnostic buffers, communications alarms on the HMI, network latency and the passive sensor's alerts. The SOC confirms it can see the test traffic, which doubles as a detection test.
- After the test: compare controller logic and mode with pre-test state to confirm nothing changed.
Metrics that show the programme is safe and useful:
- Number of process-impacting events caused by testing (target zero), and time to recover from each.
- Percentage of device families with a lab-proven profile.
- Percentage of inventory attributes (model, firmware) obtained passively versus actively.
- Number of active tests run without a signed rules of engagement (target zero).
- Percentage of test traffic detected and correctly attributed by the SOC.
Framework mapping
| Framework | Reference | How it applies |
|---|---|---|
| NIST CSF 2.0 | ID.AM-01, ID.RA-01, ID.IM-02, DE.CM-01 | Inventory, vulnerability identification, improvement from tests, network monitoring |
| NIST SP 800-53 Rev. 5 | RA-5, RA-5(5), CA-2, CA-8, CA-8(1), CM-8, CM-8(3) | Scanning, privileged scanning access, assessments, penetration testing, inventory |
| NIST SP 800-82 Rev. 3 | Appendix E network monitoring (passive and active scanning) | OT-specific guidance on choosing scanning methods |
| NIST SP 800-115 | Planning and rules of engagement | Structure for authorised technical testing |
| MITRE ATT&CK for ICS | T0846, T0846.001, T0888, T0814, T0816 | Discovery behaviours the test resembles and failure modes it can cause |
Common failure patterns
- IT scan policy applied to OT subnets: an enterprise scanner's scheduled job includes OT ranges because nobody excluded them.
- Subnet sweeps: scanning a whole range reaches devices nobody knew existed and had no tested profile.
- Fingerprinting Level 1: service detection payloads crash older communication modules.
- No operations sign-off: the control room sees communication alarms and cannot tell whether it is a test or an attack.
- Profiles never re-proven: a firmware upgrade changes the device's tolerance and the old profile now causes faults.
- Lost scanner credentials: permanent scanning accounts with broad access become an attacker's ready-made foothold.
Practitioner checklist
- Exclude OT ranges from all enterprise scan schedules by default.
- Build the inventory passively first and document which gaps need active queries.
- Write a rules of engagement template with scope, approvals, stop criteria and recovery plan.
- Maintain a lab of matching models and firmware and prove every profile there.
- Throttle scans, use known address and port lists, and disable fingerprinting on Level 0 to 1.
- Exclude safety systems and protection relays from production active testing.
- Issue scanner credentials per engagement from a vault and disable them afterwards.
- Notify the SOC and control room, and confirm the test is visible to monitoring.
- Verify controller state after testing and record any harm in the profile library.
Standards and references
- NIST: SP 800-82 Rev. 3 Guide to Operational Technology (OT) Security
- NIST: SP 800-115 Technical Guide to Information Security Testing and Assessment
- NIST: SP 800-53 Rev. 5
- NIST: Cybersecurity Framework 2.0
- CISA: Foundations for OT Cybersecurity, Asset Inventory Guidance for Owners and Operators
- CISA: APT Cyber Tools Targeting ICS/SCADA Devices (AA22-103A)
- MITRE ATT&CK: Remote System Discovery (T0846)
- Nmap: Timing Templates
Related NHI Mgmt Group resources: OT Asset Visibility Guide · OT Penetration Testing Guide · ICS Advisory Triage and OT Vulnerability Management Guide · OT Network Security Monitoring Guide