High-voltage electrical substation with steel structures and power lines at dawn

Passive PQC Discovery in OT: Finding Cryptography Safely

Most PQC discovery advice starts with a scan. Point a tool at your address ranges, let it complete a handshake with every listening service, and record the protocol versions, cipher suites, key exchange groups and certificates it finds. On an office network that is routine. On a control network it can upset a controller that was never built to answer unexpected connections.

Operational technology is also where a cryptographic inventory takes longest to act on. The UK NCSC’s timelines for migration to post-quantum cryptography are aimed in part at “operators of critical national infrastructure systems including industrial control systems”, and set 2028 for discovery and an initial plan, 2031 for priority migrations and 2035 for completion. The same guidance warns that industrial IoT devices “might not be upgradeable” and “might not be designed to be replaced”. NIST’s final IR 8547 points the same way: quantum-vulnerable public-key algorithms such as RSA and ECDSA, at the key sizes in common use, are deprecated after 2030 and disallowed after 2035.

This article explains how to carry out passive PQC discovery in OT: what traffic alone can tell you, how to collect it without touching devices, what it will miss, and how the results end up in an inventory you can plan from.

Why active scanning is a risk on control networks

The official guidance is consistent. NIST SP 800-82 Rev. 3, the Guide to Operational Technology (OT) Security, says owners “should exercise extreme caution when permitting active scanning on an operational network due to device sensitivity on the target network”, because active scans “may cause device instability or interfere with the device process state, potentially impacting safety and integrity.” It recommends scheduling them “during planned OT outages whenever possible” and testing automated asset tools on offline systems before they go near production. A Revision 4 initial public draft was published on 21 September 2026 for comment until 30 November 2026; Rev. 3 remains the final version.

The NCSC’s 2025 guidance on creating and maintaining a definitive view of your OT architecture is blunter. Active scanning “has the potential to overwhelm legacy devices that may have limited processing power, leading to performance degradation, freezing or crashing.” If you do scan, it asks for native ICS/OT protocols and tools, methods “tested and validated by the original equipment manufacturer (OEM) of each asset”, advance notice to your SOC and planned maintenance windows. Its list of information sources puts passive monitoring well ahead of active scanning: “Use passive monitoring tools to help build your OT architecture record.”

The joint CISA, EPA, NSA, FBI and international partner guide Foundations for OT Cybersecurity: Asset Inventory Guidance for Owners and Operators (13 August 2025) does not deal with cryptography as such, but its high-priority asset attributes include “Active/supported communication protocols” and “Ports/services”. Those are the fields a cryptographic inventory builds on.

What a passive sensor can see on the wire

Encryption covers the payload. The negotiation that sets it up happens before any session keys exist, and a sensor watching a copy of the traffic can read it. NIST notes that passive scanning cannot “inspect encrypted traffic (without special provisioning)”. For PQC discovery that matters less than it sounds, because the parameters you need are exchanged first.

TLS. The ClientHello offers cipher suites and key exchange groups and the ServerHello records the server’s choice; neither is encrypted. TLS 1.3 is now specified in RFC 9846 (July 2026), which obsoletes RFC 8446. It states that “All handshake messages after the ServerHello are now encrypted”, and its handshake diagram shows the server’s Certificate protected with handshake keys. In TLS 1.2 the certificate travels before encryption starts. So a passive sensor can read server certificates on TLS 1.2 sessions but not on TLS 1.3. Post-quantum key exchange, which is defined for TLS 1.3, stays visible: RFC 10024 (August 2026) defines hybrid groups such as X25519MLKEM768, which the IANA TLS Supported Groups registry lists as value 4588.

SSH. Under RFC 4253, each side starts key exchange with an SSH_MSG_KEXINIT message listing its key exchange, host key, encryption and MAC algorithms in order of preference, and the new keys only take effect after SSH_MSG_NEWKEYS. Both offers are readable, and the RFC’s selection rule tells you which key exchange method was chosen. OpenSSH 10.0 (April 2025) made the hybrid mlkem768x25519-sha256 its default, so SSH into engineering hosts is a quick place to see post-quantum adoption, or its absence.

IPsec. In IKEv2 (RFC 7296) the first pair of messages, IKE_SA_INIT, “negotiate cryptographic algorithms, exchange nonces, and do a Diffie-Hellman exchange”. Identities and certificates follow in IKE_AUTH inside an encrypted payload, so they are hidden from a passive observer. RFC 9370 lets IKEv2 run additional key exchanges alongside the first, which is how post-quantum methods such as ML-KEM are being added, and those extra key exchange transforms are proposed and selected in the IKE_SA_INIT SA payload, so they are visible too.

OPC UA. OPC UA Secure Conversation, defined in OPC 10000-6, secures OpenSecureChannel messages with asymmetric cryptography. Their security header carries the SecurityPolicyUri, “the URI of the Security Policy used to secure the Message”, and the sender’s certificate, and it comes before the sequence header that the specification describes as the start of “the first encrypted block”. A sensor that decodes OPC UA can therefore usually tell which policy a channel uses. The specification says the mechanism “is designed for use with asymmetric cryptography algorithms, such as RSA”, with ECC variants using Diffie-Hellman key negotiation. Both families are quantum-vulnerable.

Modbus and IEC 62351. The Modbus Organization’s Modbus Security protocol wraps Modbus in TLS with X.509v3 certificates on port 802, and IEC 62351-3:2023 profiles TLS 1.2 and TLS 1.3 for power-system protocols over TCP/IP. Both inherit the TLS rules above, including the hidden certificate under TLS 1.3. For any other industrial protocol with its own security extension, check your sensor’s protocol support before relying on passive results.

ProtocolVisible to a passive sensorNot visible
TLS 1.2Version, offered and chosen cipher suites, key exchange groups, server certificateApplication data
TLS 1.3Offered and chosen cipher suite and key exchange group, including hybrid ML-KEM groupsServer and client certificates
SSHBoth sides’ algorithm lists from SSH_MSG_KEXINITSession contents
IKEv2Proposals and key exchange methods in IKE_SA_INITEncrypted parts of IKE_AUTH, including certificates
OPC UASecurityPolicyUri and sender certificate in OpenSecureChannelEncrypted message bodies

One limit applies to all of them. A sensor records what was negotiated, not everything a device would accept. A server’s weakest supported option only shows up if a client asks for it.

Collecting traffic without touching devices

Passive collection means a copy of the traffic reaches the sensor and nothing goes back. The two usual sources are SPAN (mirror) ports on managed switches and network TAPs. Gigamon describes its network taps as “More reliable than SPAN” and says its active and passive taps “are unidirectional, with data flowing strictly from the network to your tools.” Larger sites often add a packet broker so that many capture points can feed several tools.

Fitting a tap is still a physical change. NIST SP 800-82 asks organisations to consider the impact of network monitoring tools “(e.g., in-line network probes)” on OT systems, including “the momentary disconnection of manufacturing system components from the network”, and to use planned downtime. A SPAN session is lighter, but it is still a configuration change to a production switch and needs the same approval.

Sensor placement decides what you see. Start with the conduits between zones, the IT/OT DMZ and the remote-access paths into the control network. The NCSC singles out remote logins into ICS IT zones as channels that “will need to become quantum-secure”.

Two of our partners work at this layer. Nozomi Networks’ Guardian sensor “passively observes and analyzes local network traffic”, sits “on mirrored network ports or taps”, and builds an asset inventory with device type, firmware version and serial number. It offers a passive-only mode for environments that “prohibit active querying”. Its product page does not describe a cryptographic inventory, so use it for asset and protocol context: which devices talk, over what, to whom. Gigamon’s PQC readiness summary goes further on cryptography. It says the platform “Reveals cryptographic key exchange, including ML-KEM, and TLS and SSH protocol versions”, “Classifies traffic as PQC Safe, PQC Vulnerable, or Migration Needed”, works “without decryption” and covers “OT and IoT”. Gigamon announced PQC support in GigaVUE 6.12 on 5 November 2025.

What passive discovery cannot see

Passive discovery is a safe first pass. Its gaps are predictable.

  • Quiet devices and rare sessions. NIST says passive scanning “cannot identify devices that are not actively communicating” and “will often take days to complete”. Vendor support sessions, periodic engineering downloads and certificate renewals may not happen during your capture window.
  • Certificates under TLS 1.3 and in IKE_AUTH. Key algorithm, size and expiry of those certificates have to come from the issuing CA or the device configuration. Our guide to PKI and HSM discovery covers that side.
  • Cryptography that never crosses the network. Firmware signing keys, keys held in HSMs or secure elements, and encrypted data at rest leave no trace in traffic.
  • Non-IP links. NIST notes that automated topology discovery “can only capture details from IP-based networked devices”. Serial and fieldbus segments need another method.

CISA’s strategy for automated cryptography discovery and inventory (15 August 2024) says the same of automation in general: “software packages with embedded cryptography, customized applications, or certain operating systems may not be discoverable using automated tools.”

Closing the gaps safely

  1. Ask the manufacturer. The NCSC notes that a Software Bill of Materials and Hardware Bill of Materials can be requested from manufacturers. Ask OEMs which algorithms each product uses, where keys are stored, and whether cryptography can be changed by a firmware update or only by replacement.
  2. Use the PKI as a source. The CA that issues OT certificates holds the record that TLS 1.3 hides from the wire.
  3. Review engineering workstations, historians and OT servers. These hosts often run general-purpose operating systems and can be assessed with read-only, credentialed methods agreed with the asset owner, without probing field devices.
  4. Save active checks for maintenance windows. Follow the NCSC conditions above, and test the tool offline first, as NIST recommends.
  5. Inspect what cannot be observed. The CISA asset inventory guide tells owners to identify assets through “a physical inspection and logical survey”. Physical inspection is still how serial devices and isolated systems get into the record.

Record how each finding was obtained. A Cryptographic Bill of Materials that shows whether an entry came from traffic, configuration or vendor evidence is easier to trust and to keep current.

Where Quantum Sentinel fits

Quantum Sentinel is an enterprise cryptographic exposure management platform and one of our technology partners. In OT it matters exactly how a tool collects data, so here is what its own pages say.

Collection routes. The platform page lists “active TLS, SSH and IPsec scanning, an optional passive sensor on a mirror port, and read-only connectors” to HSMs, PKI, key managers, network appliances, databases and other systems. The pages we reviewed do not describe what the passive sensor records or which industrial protocols it decodes, so confirm that against your own traffic in a pilot. Quantum Sentinel also states that “Traffic is never decrypted, intercepted or proxied.”

Active checks under control. Its cryptographic coverage page says OT and IoT are assessed “under profiles built for fragile equipment: non-disruptive by default, with low-and-slow pacing available and workers deployed inside the segment”, and that probes that could destabilise OT equipment “are withheld unless explicitly authorised, and what was withheld is recorded with the scan.” Scan workers “pull work outbound, so no inbound path into a protected segment is needed”, and the platform installs in air-gapped environments.

One inventory for IT and OT. For critical infrastructure, Quantum Sentinel describes “A living inventory spanning IT and OT, from control-room systems to field devices”, with an exposure score used to “Flag hardware-embedded keys that are hardest to rotate”. Each CBOM entry records the source that found it with first-seen and last-seen timestamps, a quantum risk score and a harvest-now-decrypt-later flag, and the inventory exports as CycloneDX 1.7. One boundary to plan around: the CBOM page states that this version does not ingest third-party SBOM or CBOM inventories, although the coverage page says findings from third-party scanners and existing inventories are re-assessed against the same classification. Confirm with Quantum Sentinel early how findings from other passive tools will be brought into the inventory.

How Sinevis Can Help

Sinevis plans and runs OT cryptographic discovery that starts passive and stays within your operational constraints. We work with Nozomi Networks for OT and IoT asset visibility, Gigamon for traffic access and encrypted-traffic metadata, and Quantum Sentinel for the cryptographic inventory and exposure assessment. All three are among our technology partners. If harvest-now-decrypt-later is the starting concern, our explainer on post-quantum threats sets out the background.

To scope a passive discovery exercise or a wider post-quantum readiness assessment, contact the Sinevis team.

Sources