Manufacturing & OT
Operational Technology (OT) demands a totally different security mindset. Unlike IT environments where confidentiality reigns supreme, OT prioritizes availability and safety above all else. A compromised database is a PR nightmare; a compromised robotic arm can cause kinetic damage and halt millions of dollars in production.
This lesson is AI-generated and reviewed by a practitioner before publication. That review checks for accuracy and usefulness — it is not a formal audit, and no external body has certified this material.
Revised means the page changed on that date — not that every fact on it was re-verified then. Where an individual claim has been checked against a primary source, a "Verified" note appears next to it.
What that means for you: treat framework and regulatory detail here (GDPR, NIS2, the EU AI Act, NERC CIP, CMMC, HIPAA, PCI DSS and the rest) as a well-informed starting point, not as authority. Regulations change and AI-assisted text can state stale or subtly wrong specifics with complete confidence. Before you act on a control mapping, an obligation, or a deadline — especially in a filing, an audit response, or a board paper — verify it against the primary source. Where a lesson cites a date or a vendor, look for the "Verified" note next to it.
Real-World Example means a documented, publicly reported incident you can look up and verify independently. Illustrative Scenario means a composite drawn from patterns that recur across engagements — the dynamics and the lesson are real, but the organisation is not a specific named company and the figures are representative rather than audited. We label them differently so you always know which is which, and never cite a composite to your board as though it were a documented case.
IT/OT Convergence
Historically, Information Technology (IT) and Operational Technology (OT) were entirely separate domains. IT handled emails and spreadsheets; OT handled valves, PLCs (Programmable Logic Controllers), and assembly lines. Today, the demand for data analytics, remote monitoring, and edge-to-cloud efficiency has forced these domains to merge. This is not a gradual evolution — it is a collision, and the blast radius lands squarely on the CISO's desk.
The Culture Clash
The role of the CISO in a manufacturing environment is often diplomatic. You are bridging a cultural divide. IT professionals prioritize Confidentiality, Integrity, and Availability (the CIA triad) — usually in that exact order. OT engineers prioritize Availability, Reliability, and Safety (ARS). Asking an OT plant manager to patch a legacy Windows XP PLC controller because of an IT vulnerability scan will be met with outright hostility if it requires shutting down the line.
The root of the conflict is economic reality. An automotive stamping line running at capacity generates $50,000–$200,000 per hour. A patch window that takes a production line offline for four hours costs more than most annual security budgets. OT engineers are not being obstructionist — they are doing their jobs. Your job is to find security outcomes that do not require them to choose between uptime and security posture. That requires understanding their constraints before proposing solutions.
| Dimension | IT Priority | OT Priority |
|---|---|---|
| Primary goal | Data confidentiality | Physical process continuity |
| Patch cadence | Monthly (Patch Tuesday) | Annual or per vendor certification |
| System lifecycle | 3–5 years | 15–25 years |
| Availability tolerance | Planned maintenance windows | Zero unplanned downtime |
| Security testing | Active scanning acceptable | Passive-only; active scanning can crash equipment |
| Authentication model | MFA, SSO, directory services | Shared accounts, local credentials common |
| Change management | Agile / continuous deployment | Change freeze periods; vendor-validated changes only |
| Incident response | Isolate and contain quickly | Graceful shutdown; safety interlock sequencing required |
Never scan actively. Standard IT vulnerability scanners (like Nessus or Qualys) will often crash legacy OT equipment just by pinging it or running port scans. OT discovery must be done passively via SPAN ports or TAP devices listening to network traffic. This is not a suggestion — it is a hard operational constraint.
OT-Specific Monitoring Tools
Standard SIEM and EDR tools are blind to OT protocols. Modbus, DNP3, EtherNet/IP, and PROFINET are not parsed by Splunk out of the box. A new category of OT-native security monitoring platforms has emerged to fill this gap. The three market leaders — Claroty, Nozomi Networks, and Dragos — all work on the passive-monitoring principle: they sit on SPAN ports, ingest raw traffic, decode OT protocols natively, and build an asset inventory and behavioral baseline without injecting a single packet into the control network.
The differentiation between these platforms matters when you are writing an RFP. Dragos is built primarily by former ICS incident responders and is strongest in threat intelligence and incident response workflow integration — it is the platform of choice when your primary concern is nation-state threat actors. Nozomi Networks has the broadest protocol support and strongest asset management features, making it the default choice for environments with heterogeneous equipment from multiple vendors across multiple generations. Claroty has the tightest integration with enterprise IT tooling (ServiceNow, Splunk, Palo Alto Cortex) which makes it the preference when your primary pain point is breaking down the IT/OT operational silo. All three are viable; the right choice depends on your threat model and integration requirements, not marketing claims.
LockerGoga ransomware encrypted systems across Norsk Hydro's global aluminum production operations. The company deliberately chose to run manual operations rather than pay ransom, absorbing an estimated $71 million USD in losses. Recovery took weeks. Post-incident analysis showed the initial access was through IT systems — the convergence path from IT to OT gave the ransomware its reach across production. Norsk Hydro has since been transparent about their recovery and is now considered a model for OT resilience planning.
Governance Structures for Converged Environments
Technology controls alone will not fix IT/OT convergence security. You need a governance structure that gives both domains a legitimate voice and assigns clear accountability. The most effective model is a joint IT/OT Security Committee that meets monthly and includes the CISO, the OT/plant engineering lead, the network operations lead, and a representative from physical security. This committee owns the converged security roadmap, reviews incidents that crossed the IT/OT boundary, and approves any changes to the IDMZ architecture.
Below that committee, a RACI matrix is essential for clarifying who owns what at the operational level. Without it, you will find that nobody owns the historian server, nobody owns the IDMZ firewall rules, and nobody owns contractor access management — until something goes wrong. A minimal RACI for a converged environment should cover: asset inventory ownership, vulnerability disclosure and patch coordination, OT-device onboarding approval, vendor/contractor access provisioning, and security incident declaration authority.
The Purdue Model
The Purdue Enterprise Reference Architecture (PERA), developed in the 1990s, is the gold standard for segmenting manufacturing networks. It structures the enterprise into distinct levels, creating defensible boundaries between business functions and industrial control systems (ICS). If you are a CISO at a manufacturer and you cannot draw this architecture from memory, you are missing the foundational vocabulary for every OT security conversation you will have.
The Six Levels
| Level | Name | Function | Primary Security Control |
|---|---|---|---|
| Level 5 | Enterprise Network | Corporate IT, email, web access. | Standard IT controls (EDR, email gateway, PAM) |
| Level 4 | Site Business Planning | Local IT, ERP systems, local corporate services. | Network segmentation from Level 5; strict routing policy |
| Level 3.5 | Industrial DMZ (IDMZ) | The critical buffer zone. Jump servers, proxy servers, data brokers. | The defensive chokepoint. All cross-zone traffic terminates here. |
| Level 3 | Site Manufacturing Ops | Plant-wide control systems, OSIsoft PI/AVEVA historians. | Application whitelisting; OT-native monitoring (Nozomi/Claroty) |
| Level 2 | Area Supervisory Control | HMI (Human Machine Interfaces), supervisory control rooms. | Physical access controls; no internet-routable addresses |
| Level 1 | Basic Control | PLCs, RTUs, intelligent controllers. | Protocol filtering; firmware integrity monitoring |
| Level 0 | Process | The actual physical equipment: sensors, valves, motors. | Physical security; safety instrumented systems (SIS) |
As a CISO, your primary architectural goal is enforcing strict traffic controls at the Level 3.5 IDMZ. No traffic should ever flow directly from Level 4/5 down to Level 2/1. All communication must terminate in the DMZ and be re-brokered downwards by an application-layer proxy. This is not a theoretical ideal — it is the only architectural pattern that gives you a defensible chokepoint. Without it, a compromised ERP server becomes a direct path to your PLCs.
In 2021, a water treatment facility in Oldsmar, Florida was remotely accessed via TeamViewer and an operator watched as an unknown actor attempted to raise sodium hydroxide levels to dangerous concentrations. Investigation revealed the facility's IT and OT networks were on the same flat network segment with no Purdue-model segmentation. The attacker's path from internet to process control was direct and unobstructed. The only thing that stopped a public health incident was an alert operator physically watching the screen.
Implementing Network Segmentation
Retrofitting Purdue segmentation into a brownfield manufacturing environment is a multi-year program, not a project. Prioritize ruthlessly. Start with the IDMZ: if you have none, creating one is your first deliverable. Deploy next-generation firewalls at the Level 3/3.5 boundary with strict default-deny policies and log everything. Use VLANs within each level to limit lateral movement. For the most sensitive Level 1/0 environments, consider unidirectional security gateways — hardware-enforced data diodes from vendors like Waterfall Security Solutions or Owl Cyber Defense that are physically incapable of allowing traffic to flow upward from OT to IT.
Practical sequencing for brownfield environments: First, do a full passive asset discovery using an OT monitoring platform — you cannot segment what you have not inventoried. Second, document all data flows that legitimately cross level boundaries (historian pulls, remote access sessions, patch delivery paths) before you put any firewall rules in place. Cutting legitimate flows without understanding them will stop production and destroy your credibility. Third, implement deny rules incrementally with a rollback plan for each change. Fourth, implement Just-In-Time (JIT) privileged access for any remote access that crosses the IDMZ rather than persistent VPN tunnels.
While the Purdue model gives you the architecture, IEC 62443 gives you the governance and operational framework. It defines Security Levels (SL 1–4) for zones and conduits, which map directly to how strictly you control traffic at each Purdue boundary. IEC 62443-3-3 covers system security requirements; 62443-2-1 covers the security management system. It is the de-facto standard for industrial automation and control system (IACS) security and the basis for most OT-relevant regulatory frameworks globally.
Modernizing Purdue for Cloud and IIoT
The Purdue model was designed for a world where all data stayed on-premises and all devices were physically wired. The Industrial Internet of Things (IIoT) and cloud historians have broken both assumptions. Wireless sensors reporting directly to Azure IoT Hub or AWS IoT Core exist outside the Purdue hierarchy entirely. Cloud-based historians like AVEVA Connect or OSIsoft PI Cloud bypass the IDMZ by design. This does not mean Purdue is obsolete — it means you need an extended architecture that accounts for these new pathways.
The emerging answer is the ISA/IEC 62443 Zone and Conduit model used alongside a cloud-aware variant of Purdue. Every data path to the cloud must be treated as a conduit and assigned a Security Level. IoT devices should communicate through a local industrial IoT gateway (a DMZ equivalent) rather than directly to the cloud, so you retain a chokepoint. Zero Trust principles apply to cloud-bound OT data: authenticate devices, encrypt all telemetry in transit, and enforce data classification policies that prevent raw process data from being accessible to unauthorized cloud tenants. The architecture is more complex than classic Purdue, but the underlying principle is identical: no direct path from the physical process to an untrusted network.
The Air-Gap Myth
The most common pushback you will hear from plant managers is: "Our systems are secure because they are air-gapped." This is almost universally false in modern manufacturing. True air-gaps — zero physical or logical connection of any kind — exist only in the most classified military and nuclear contexts, and even there, attackers have found ways across. In commercial manufacturing, the air-gap is a story people tell themselves to avoid having a harder conversation about security investment.
How the Gap is Bridged
Production lines require maintenance, updates, and data extraction. Every one of those operational needs creates a potential bridge across any supposed air-gap. The gap is broken daily by contractors bringing in unmanaged laptops to tune PLCs, by USB drives used to transfer patch files or pull telemetry data, by undocumented LTE modems installed by engineers for remote access convenience, and by legitimate remote access connections that were put in temporarily and never removed. Ask a plant manager to produce a complete list of every device that has connected to the OT network in the past 12 months. Almost none of them can.
The most famous cyberweapon in history, Stuxnet, successfully compromised the Natanz nuclear facility — an environment designed to be genuinely air-gapped. The initial vector was targeted delivery via corrupted USB thumb drives carried in by internal engineers. If nation-state adversaries can bridge a real air-gap at an Iranian nuclear enrichment facility, assume your plant's "air-gap" provides no meaningful protection against a determined attacker.
Assessing Your Real Attack Surface
The first step in dispelling the air-gap myth within your own organization is conducting a physical and logical gap analysis. This is a structured process, not an IT assessment — it requires walking the plant floor. Physical gap analysis: identify every device connected to the control network by physically tracing cables and documenting every connection point. Look for rogue wireless access points, cellular modems, and undocumented serial connections. This cannot be done from a network diagram — those are almost always wrong. Logical gap analysis: deploy passive OT monitoring and let it run for 30 days before you review results. Every device it discovers that is not on your asset register is a gap-bridging candidate.
Document your findings as an OT attack surface register, not just a vulnerability list. For each bridging vector you find, document: the device or connection type, who installed it and when, what legitimate business purpose it serves, whether it is actively monitored, and what the blast radius would be if it were compromised. This register is your evidence base for security investment conversations with plant management and with the board. Abstract risk statements do not move budget; a list of thirty undocumented modems connected to your Level 2 network does.
| Air-Gap Bridging Vector | How Common | Detection Method | Mitigation Strategy |
|---|---|---|---|
| Contractor laptops / portable engineering workstations | Very high | OT asset monitoring; visitor log cross-reference | Contractor device policy; mandatory scan at USB kiosk before connection |
| USB drives (patch delivery, data extraction) | Very high | USB activity logging on hosts; DLP on engineering workstations | USB scanning kiosks with malware inspection; removable media policy; hardware-level blocking on non-essential systems |
| Rogue LTE / cellular modems | High | RF spectrum scanning; OT network traffic anomaly detection | Physical inspection program; cellular signal detection tools; strict change management |
| Persistent remote access (forgotten VPN / TeamViewer) | High | Network flow analysis; firewall rule audit | Quarterly remote access review; Just-In-Time access; session recording |
| Historian / data aggregation servers with dual-NIC | Medium | Network diagram audit; asset discovery | Remove dual-homing; route all historian traffic through IDMZ proxy |
| Wireless sensor networks (IIoT devices) | Growing | Wireless spectrum monitoring; OT asset inventory | Isolated VLAN; IoT gateway as enforcement point; certificate-based auth |
| Vendor-installed diagnostic ports (undocumented) | Medium | Physical inspection; vendor contract review | Contractual disclosure requirement; physical port covers; commissioning audit |
Compensating Controls for Semi-Connected Environments
The practical reality is that most manufacturing environments are neither fully connected nor truly air-gapped — they are semi-connected, with controlled but real pathways across the boundary. The security objective is not to eliminate these pathways but to make them controlled, logged, and inspected. Three compensating controls are essential for any semi-connected OT environment.
First, USB scanning kiosks: deploy locked-down workstations running purpose-built malware scanning software (vendors include Sasa Software, OPSWAT, and Honeywell's Secure Media Exchange) at all physical entry points to the plant. Any removable media must be scanned before it is used on any OT system. This directly addresses the most common gap-bridging vector without requiring you to eliminate USB usage — which is operationally impractical in most plants. Second, application whitelisting on engineering workstations: tools like Tripwire or Carbon Black App Control enforce a default-deny execution policy so that only approved binaries can run. A contractor's compromised laptop cannot execute malware if every unknown binary is blocked by default. Third, data diodes for one-way data flows: for environments where data must flow outbound (telemetry, historian data) but nothing should flow inbound, hardware data diodes from Waterfall Security or Owl Cyber Defense are the correct solution. They are not firewalls — they are hardware devices that are physically incapable of carrying traffic in the wrong direction.
Stop trying to prove the air-gap exists. Start building a security model that assumes the air-gap does not exist and implements layered controls accordingly. Your threat model should assume an adversary can reach Level 3 through some vector — then ask what stops them reaching Level 1/0. That is the question your architecture needs to answer.
Supply Chain & Vendor Risk
Manufacturing supply chains are complex and deeply intertwined. The classic axiom — "you are only as secure as your weakest supplier" — understates the problem. It is not just that a weak supplier creates a vulnerability; it is that a supplier with access to your OT environment can become the attack vector that bypasses every perimeter control you have ever built. If an attacker cannot breach your network directly, they will target your automation vendor, your remote monitoring provider, or the contractor who services your HVAC systems.
Vendor Risk Classification
A CISO cannot personally audit hundreds of vendors. Implement a tiered vendor risk management (VRM) program with a classification model that forces prioritization. Classify vendors based on two primary vectors: OT network access (does this vendor have persistent or periodic access to Level 2/3 environments?) and business criticality (if this vendor is unavailable or compromised for 72 hours, does production halt?). The intersection of these two vectors determines the tier and the associated security requirements.
| Tier | OT Network Access | Business Criticality | Required Controls | Assessment Frequency |
|---|---|---|---|---|
| Tier 1 — Critical | Persistent or frequent OT access | Production-halting if unavailable | SOC 2 Type II; annual on-site audit; JIT-only access; session recording; incident notification SLA <4h | Annual + event-triggered |
| Tier 2 — High | Occasional OT access or IT access with OT data | Significant operational impact | SOC 2 Type II or equivalent attestation; penetration test summary; MFA required; access reviewed quarterly | Annual |
| Tier 3 — Medium | IT-only access; no OT network path | Moderate impact; workarounds available | Security questionnaire (SIG Lite or equivalent); MFA; incident notification SLA <24h | Every 2 years |
| Tier 4 — Low | No network access; commercial relationship only | Minimal operational impact | Standard contract security clauses; self-attestation | As needed |
Tier 1 vendors should never have persistent VPN credentials into your OT network. Every access session must be provisioned on-demand through a PAM platform (CyberArk, BeyondTrust), time-limited, scoped to specific assets, and fully recorded with video capture. Session credentials are rotated after every use. This eliminates the standing access that made the Target breach (via HVAC vendor) and the Colonial Pipeline breach possible.
Contractual Security Requirements
The security questionnaire is not enough. For Tier 1 and Tier 2 OT vendors, your contracts must contain enforceable security obligations. Most procurement teams will resist adding security clauses because it slows deals — your job is to make the legal and compliance case for why these clauses are non-negotiable for vendors with OT access. Key clauses to require in OT vendor contracts include: right to audit (the right to conduct or commission a security audit of the vendor's systems with 30 days' notice), incident notification SLA (contractual obligation to notify you within 4 hours of any confirmed or suspected breach that could affect your environment), sub-contractor flow-down (all security obligations must flow down to sub-contractors and fourth parties with equivalent requirements), vulnerability disclosure (vendor must disclose known vulnerabilities in products or services they deliver within 10 business days of discovery), and access device standards (vendor technicians must use managed, patched devices that meet your endpoint security requirements when connecting to your OT network).
These clauses have teeth only if you enforce them. Build a vendor security review into your annual contract renewal process. For Tier 1 vendors, conduct a tabletop exercise once per year that includes the vendor's security team alongside yours — if they cannot respond coherently to a simulated breach scenario, that is material information about whether their security posture matches their contractual attestations.
If you operate in regulated sectors, supply chain security requirements are increasingly mandatory rather than advisory. NERC CIP (energy), NIS2 (EU), and the forthcoming EU Cyber Resilience Act all contain supply chain security provisions. CMMC 2.0 (US defense contractors) requires that supply chain risk management controls flow down to sub-contractors. Treat your VRM program as a regulatory compliance foundation, not just a security best practice.
Incident Scenarios: When a Vendor Gets Breached
The SolarWinds supply chain attack in 2020 demonstrated what a sophisticated vendor compromise looks like at scale: a malicious update pushed through a trusted software vendor's own update mechanism, signed with the vendor's legitimate certificates, delivered to approximately 18,000 organizations globally including critical infrastructure operators. In a manufacturing context, the equivalent scenario is your automation vendor's remote access software or your SCADA platform's update mechanism being compromised. Your controls at the perimeter are irrelevant — the malicious payload arrives via a trusted, cryptographically-signed update channel.
NotPetya (2017) provides the most financially devastating supply chain attack example in manufacturing history. Disguised as ransomware but actually a destructive wiper, NotPetya was initially delivered through a compromised Ukrainian accounting software update (M.E.Doc). It spread globally through networks using a weaponized EternalBlue exploit. Maersk, the world's largest shipping company, lost an estimated $300 million and had to reinstall 45,000 PCs, 4,000 servers, and 2,500 applications in 10 days. Merck, the pharmaceutical manufacturer, lost an estimated $870 million. Mondelez lost $188 million. FedEx's TNT subsidiary lost $400 million. These were not small companies with immature security programs — they were global enterprises with security teams and budgets. The lesson is not that their security failed; it is that supply chain attacks can bypass conventional security controls entirely.
The practical implication for manufacturing CISOs is that vendor risk management must include supply chain attack scenarios in your incident response planning. Test your ability to isolate and recover OT systems without the vendor's remote access tools if those tools are compromised. Maintain offline backups of all OT system configurations and validate them quarterly. Implement network-level controls that restrict what even trusted vendor update channels can access on your OT network — a software update mechanism does not need to communicate with your Level 1 PLCs. Defense in depth applies to trusted vendors just as much as it applies to adversaries.
The REvil ransomware group exploited a zero-day in Kaseya's VSA remote monitoring and management platform, pushing malicious updates to approximately 1,500 managed service provider (MSP) customers simultaneously. For manufacturing companies using MSPs for IT support, this was a cascading supply chain attack: the attacker breached the MSP's tooling to reach the MSP's clients. REvil demanded $70 million for a universal decryption key. This scenario is directly applicable to any manufacturer using an MSP or managed security service provider with remote management access to your environment.
ICS/SCADA Protocol Security
Industrial control systems communicate using protocols that were designed decades ago for reliability and determinism — not security. Understanding the protocol landscape is essential for any CISO operating in a manufacturing environment, because the vulnerabilities are not implementation bugs that can be patched. They are architectural design decisions baked into the protocol specifications themselves. You cannot firewall your way out of a protocol that has no concept of authentication.
The Major OT Protocols
| Protocol | Primary Use | Authentication | Encryption | Key Vulnerability |
|---|---|---|---|---|
| Modbus TCP/RTU | Register read/write for PLCs, sensors | None by design | None | Any device on the network can issue commands to any Modbus device |
| DNP3 | Utilities, water treatment, some manufacturing SCADA | Optional (Secure Authentication v5) | None natively | Unauthenticated by default; SA v5 adoption is minimal |
| OPC Classic (DA/HDA) | Data exchange between HMIs, historians, SCADA | DCOM-based (Windows auth) | None | Relies on DCOM — massive attack surface; firewall-hostile |
| OPC UA | Modern replacement for OPC Classic | X.509 certificates, user tokens | TLS 1.2+ | Secure by design but misconfigured deployments common |
| EtherNet/IP (CIP) | Rockwell/Allen-Bradley automation | None natively | None natively | CIP commands are unauthenticated; mode changes possible remotely |
| PROFINET | Siemens automation ecosystems | None natively | None natively | Real-time Ethernet; no cryptographic integrity checks on process data |
The pattern is unmistakable: the protocols that run the vast majority of manufacturing equipment worldwide — Modbus, EtherNet/IP, PROFINET — were built with zero authentication and zero encryption. This is not an oversight. When these protocols were designed, OT networks were physically isolated and the threat model did not include network-based adversaries. The convergence of IT and OT networks has placed these protocols in an environment they were never designed to survive.
Protocol-Level Attack Vectors
Because Modbus has no authentication, an attacker who gains access to a Level 2 network segment can send arbitrary read and write commands to any Modbus device. They can change setpoints, disable alarms, or put controllers into stop mode — all using perfectly valid protocol commands that are indistinguishable from legitimate traffic at the packet level. This is not a theoretical concern. The TRITON/TRISIS malware, discovered in 2017, targeted Schneider Electric Triconex Safety Instrumented Systems at a petrochemical facility in Saudi Arabia. The attackers used the proprietary TriStation protocol to reprogram the safety controllers — the systems specifically designed to prevent catastrophic physical outcomes. TRITON represents the first publicly known malware designed to disable safety systems, and it succeeded in reaching the safety controllers precisely because the protocol layer offered no resistance.
TRITON targeted Schneider Electric Triconex SIS controllers at a Saudi Arabian petrochemical plant. The attackers — attributed to a Russian government research institute (TsNIIKhM) — gained access to the SIS engineering workstation and used the TriStation protocol to deploy custom malware to the safety controllers. The attack was only discovered because a programming error caused the SIS to trip, shutting down the plant. Without that error, the attackers could have disabled safety systems while simultaneously manipulating the process — a scenario that could cause explosions, toxic releases, or loss of life.
Deep Packet Inspection and Protocol-Aware Firewalls
Standard IT firewalls see OT protocols as opaque TCP streams. A Modbus write command that changes a reactor temperature setpoint to a dangerous value passes through a standard firewall as a perfectly normal TCP packet on port 502. Protocol-aware firewalls — also called OT-aware or industrial firewalls — from vendors like Fortinet (FortiGate with OT protocol support), Palo Alto Networks (with industrial protocol App-IDs), and purpose-built solutions like Tofino (now Belden) parse the OT protocol payload and enforce rules at the application layer. They can distinguish a Modbus read from a Modbus write, block writes to specific register ranges, and alert on function codes that should never appear in normal operations.
Deep packet inspection (DPI) for OT is the detection complement to protocol-aware firewalling. OT monitoring platforms like Dragos, Nozomi, and Claroty perform DPI on mirrored traffic to detect anomalous protocol behavior — a PLC being reprogrammed outside a maintenance window, a new Modbus master appearing on the network, or firmware download commands being issued to a safety controller. This is the detective control layer for protocol-level threats.
The Encryption Challenge
The obvious IT response — "just encrypt everything" — collides with hard physics in OT environments. Real-time control loops operate on cycle times measured in milliseconds. Adding TLS handshakes and encryption overhead to a control loop that must complete in 10ms can cause the process to fail. OPC UA solves this by design with its built-in security model, and the CIP Security extension adds TLS-based authentication and encryption to EtherNet/IP. But retrofitting encryption onto legacy Modbus or PROFINET installations is not feasible without replacing the controllers — a capital expenditure measured in millions for a large plant. The practical approach is defense-in-depth: segment aggressively, deploy protocol-aware firewalls at zone boundaries, monitor with DPI, and migrate to secure protocols (OPC UA, CIP Security) as part of the natural equipment replacement cycle.
ODVA's CIP Security extension adds TLS/DTLS encryption, device authentication via X.509 certificates, and authorization policies to the EtherNet/IP protocol stack. It is the most significant security upgrade to a major OT protocol in the last decade. Rockwell Automation's newer ControlLogix and CompactLogix controllers support CIP Security natively. For greenfield deployments or controller refresh cycles, specifying CIP Security support should be a mandatory procurement requirement.
Safety Instrumented Systems (SIS)
Safety Instrumented Systems are the last line of defense between a process malfunction and a catastrophic outcome — an explosion, a toxic release, or loss of human life. In the hierarchy of manufacturing controls, the SIS sits above everything else: if the basic process control system (BPCS) fails and the process moves toward a dangerous state, the SIS must independently detect the condition and bring the process to a safe state. This is not a cybersecurity system. It is a safety system. But after TRITON, the distinction between safety and security has become the most consequential conversation in OT.
Safety vs. Security: A Critical Distinction
Safety and security are not synonyms, and confusing them creates dangerous blind spots. Safety engineering focuses on preventing harm from accidental failures — equipment malfunctions, sensor drift, human error. Security engineering focuses on preventing harm from intentional adversarial action. Safety systems are designed assuming components fail randomly; security threats are intelligent adversaries who specifically target the weakest point. A safety analysis (HAZOP, LOPA) asks "what if this valve fails open?" A security analysis asks "what if an attacker forces this valve open while simultaneously disabling the alarm?" The second question is fundamentally different and requires a different analytical framework.
| Dimension | Safety Engineering | Security Engineering |
|---|---|---|
| Threat source | Random hardware/software failures | Intelligent, adaptive adversaries |
| Primary standard | IEC 61511 / IEC 61508 | IEC 62443 / NIST CSF |
| Risk metric | Safety Integrity Level (SIL 1–4) | Security Level (SL 1–4) |
| Design principle | Redundancy, fail-safe defaults | Defense-in-depth, least privilege |
| Testing approach | Proof testing at defined intervals | Penetration testing, red teaming |
| Change management | Management of Change (MOC) per IEC 61511 | Patch management, vulnerability response |
IEC 61511 and SIL Levels
IEC 61511 is the process-sector application standard for functional safety. It defines the Safety Instrumented System lifecycle from concept through decommissioning. The Safety Integrity Level (SIL) — ranging from SIL 1 (lowest) to SIL 4 (highest) — quantifies the reliability required of the safety function. SIL 1 requires a risk reduction factor of 10–100x; SIL 3 requires 1,000–10,000x. Most manufacturing SIS functions are rated SIL 2 or SIL 3. The SIL rating determines the required hardware fault tolerance, diagnostic coverage, and proof test intervals. A SIL 3 system typically requires redundant sensors (2oo3 voting), redundant logic solvers, and proof testing every 1–2 years.
For the CISO, the critical point is that cybersecurity threats were not part of the original SIL calculations. A SIL 3-rated safety function assumes random hardware failure probabilities — it does not account for an attacker deliberately manipulating all three redundant sensors simultaneously. This is why IEC 62443 and IEC 61511 must be applied together, not independently. The 2024 revision of IEC 61511 explicitly acknowledges cybersecurity threats and requires a security risk assessment as part of the SIS lifecycle.
Why SIS Must Be Isolated from BPCS
The most critical architectural requirement for SIS is complete logical and, ideally, physical separation from the BPCS. If the same network that carries process control traffic also carries safety system traffic, an attacker who compromises the BPCS has a direct path to the SIS. This is precisely what happened in the TRITON attack — the attackers moved from the IT network to the OT network, then laterally to the SIS engineering workstation, which was on the same network as the BPCS. Best practice requires separate controllers, separate networks, separate engineering workstations, and separate credentials for the SIS and BPCS. Any integration point between the two (typically for alarm/event data) must be through a unidirectional gateway or a hardened, purpose-built interface.
Engineers often argue that SIS systems are inherently safe because they "fail safe" — meaning if they lose power or detect an internal fault, they trip to a safe state. This assumption is dangerous in a cyber context. TRITON did not cause the SIS to fail — it reprogrammed the SIS so it would not trip when it should have. An attacker who controls the SIS logic can redefine what "safe" means. The system does not fail; it operates exactly as programmed — by the attacker. Fail-safe protects against power loss and hardware faults. It does not protect against logic manipulation.
Consequence-Driven Cyber-Informed Engineering (CCE)
Idaho National Laboratory developed the CCE methodology specifically for scenarios where cyber attacks target physical processes. CCE starts not from the network or the vulnerability, but from the consequence: "What is the worst physical outcome an adversary could cause?" It then works backward through the engineering and digital pathways to identify how that outcome could be achieved and what controls would prevent it. For SIS specifically, CCE asks: "If an adversary had unlimited access and unlimited time, could they manipulate the safety system to allow a catastrophic event?" If the answer is yes — and for most networked SIS installations, it is — then the engineering design must be changed to introduce non-digital, physics-based barriers that an adversary cannot manipulate through software alone. Mechanical relief valves, rupture discs, and physical interlocks are examples of controls that exist outside the digital attack surface entirely.
After TRITON was publicly disclosed, Schneider Electric issued security advisories for Triconex controllers and released firmware updates with enhanced access control. The broader industry response was more significant: major SIS vendors (Schneider, Honeywell, ABB, Yokogawa) all accelerated development of hardened SIS architectures with built-in cybersecurity controls. NIST published SP 800-82 Rev 3 with expanded SIS guidance. The key lesson for CISOs: audit your SIS architecture now. Verify physical separation from BPCS. Verify that SIS engineering workstations are not accessible from the general OT network. Verify that SIS firmware versions are current and that default credentials have been changed.
OT Incident Response
Incident response in OT environments is fundamentally different from IT incident response, and applying IT-centric IR playbooks to OT incidents will at best waste time and at worst cause physical damage. The differences are not merely procedural — they are rooted in the physics of industrial processes, the limitations of OT forensics, and the life-safety implications of response actions. A CISO who has not developed an OT-specific IR capability is operating with a critical gap.
How OT IR Differs from IT IR
| IR Phase | IT Approach | OT Approach |
|---|---|---|
| Containment | Isolate host from network immediately | Coordinate with plant operators; unplanned isolation can cause unsafe process states |
| Evidence collection | Disk image, memory dump, log export | PLCs have no disk; limited memory; volatile ladder logic; proprietary firmware |
| Forensic tools | EnCase, FTK, Volatility | Vendor-specific engineering software; protocol captures; Dragos forensic tools |
| Recovery priority | Restore data, then services | Restore safety systems first, then process control, then data |
| Communication | IT security team, management, legal | Plant operators, safety officers, process engineers — before IT leadership |
| Regulatory reporting | Breach notification (GDPR, state laws) | Sector-specific: CISA, EPA, OSHA, state environmental agencies |
Forensics on PLCs and Controllers
Traditional digital forensics assumes you can image a hard drive, capture volatile memory, and analyze filesystem artifacts. PLCs have none of these. A PLC is a purpose-built embedded system with proprietary firmware, ladder logic stored in non-standard memory, and no filesystem in the traditional sense. Forensic analysis of a compromised PLC requires the vendor's engineering software (e.g., Rockwell's Studio 5000, Siemens TIA Portal) to download and compare the running logic against a known-good baseline. If you do not have verified backups of your PLC programs — and most organizations do not — you have no baseline to compare against. This is why OT asset management and configuration backup are prerequisites for OT incident response, not optional enhancements.
Network-based forensics are often the most productive evidence source in OT investigations. If you have an OT monitoring platform (Dragos, Nozomi, Claroty) with packet capture capability, the PCAP data from the time of the incident is invaluable. It preserves exactly what commands were sent to which controllers, what responses were received, and what protocol anomalies occurred. Without network forensics, reconstructing the attacker's actions on the OT network is often impossible.
In OT incident recovery, the sequence matters as much as the actions. Always restore safety systems first. Before any process control system is brought back online, the SIS must be verified as operational and trustworthy. Then restore the BPCS and validate process control logic against known-good backups. Only then bring supporting systems (historians, HMIs, reporting) back online. Reversing this order — restoring production before verifying safety — creates a window where the process is running without its safety net. This is where the pressure from operations leadership to "get the line running" must be firmly resisted by the CISO and the safety officer together.
The Dragos IR Methodology
Dragos, founded by former NSA and ICS-CERT analysts, has published the most widely referenced OT-specific IR methodology. Their approach emphasizes four key differences from IT IR: first, collection before containment — in OT, you gather evidence before isolating systems because the isolation itself can destroy evidence (a PLC that loses power may lose volatile state data). Second, operations coordination — every containment and eradication action is reviewed with plant operations staff before execution. Third, threat-informed recovery — restoration is guided by the specific threat actor's known TTPs, not just generic recovery procedures. Fourth, community defense — anonymized indicators are shared with the OT security community through Dragos's Neighborhood Keeper platform to protect other organizations facing the same threat actor.
Building an OT-Specific IR Playbook
Your OT IR playbook must address scenarios that do not exist in IT: a PLC program modification detected during production, a safety system communication failure, ransomware spreading from IT to OT through the IDMZ, or an unauthorized engineering workstation connection to a Level 1 device. For each scenario, the playbook must define: who is notified first (always the plant operator and safety officer before IT), what evidence must be preserved (network captures, PLC program downloads, HMI screenshots), what containment options exist that do not cause an unsafe process state, and what the regulatory notification requirements are for your specific sector and jurisdiction.
An effective OT tabletop exercise must include plant operations, not just IT security staff. Scenario example: "At 2:00 AM, your OT monitoring platform alerts on unauthorized Modbus write commands to a batch reactor temperature controller. The current temperature reading is climbing above the safe operating limit. The SIS has not triggered. Your shift operator confirms the HMI shows normal values." This scenario tests whether the team can identify a potential HMI spoofing attack, coordinate between security and operations, make the decision to manually trip the process, and preserve evidence — all under time pressure. Run these exercises quarterly, rotate scenarios, and include at least one scenario per year that involves simultaneous IT and OT incidents requiring coordinated response.
Manufacturing Regulatory Compliance
Manufacturing CISOs face a regulatory landscape that is simultaneously fragmented and rapidly expanding. Unlike financial services or healthcare, where a single dominant framework (PCI DSS, HIPAA) provides a clear compliance target, manufacturing security compliance is a matrix of overlapping standards, sector-specific regulations, and emerging legislation. The CISO who waits for a single harmonized standard will wait forever. The practical approach is building a compliance program that maps common controls across multiple standards and adapts as regulations evolve.
IEC 62443 Certification
IEC 62443 is the closest thing to a universal standard for industrial cybersecurity, and it provides the most structured certification pathway. The standard defines three certification scopes that correspond to different roles in the OT ecosystem. Component-level certification (IEC 62443-4-2) certifies that an individual product — a PLC, a switch, an HMI — meets defined security capabilities at a specified Security Level. System-level certification (IEC 62443-3-3) certifies that an integrated automation system meets security requirements as deployed. Process-level certification (IEC 62443-2-4) certifies that the system integrator's processes for designing, deploying, and maintaining the system meet security requirements. For a manufacturing CISO, the practical implication is that you should require IEC 62443 component certification from vendors in procurement specifications, pursue system-level certification for your critical production zones, and require process certification from any system integrator working on your OT environment.
| Standard / Regulation | Scope | Applicability | Key Requirement for Manufacturing |
|---|---|---|---|
| IEC 62443 | Industrial automation & control | Global; voluntary but increasingly referenced by regulators | Zone/conduit model; Security Levels 1–4; component through process certification |
| NIST Manufacturing Profile | US manufacturing sector | US; voluntary (mandatory for federal contractors) | Maps NIST CSF to manufacturing-specific risks; OT asset management emphasis |
| EU Machinery Regulation (2023/1230) | Machinery placed on EU market | EU; mandatory from Jan 2027 | Cybersecurity requirements for machinery with digital elements; CE marking impact |
| Cyber Resilience Act (CRA) | Products with digital elements | EU; mandatory from 2027 | Vulnerability handling; security updates for product lifetime; applies to industrial IoT |
| FDA 21 CFR Part 11 / Premarket Guidance | Medical device manufacturing | US; mandatory for medical devices | SBOM requirements; threat modeling; post-market vulnerability management |
| CMMC 2.0 / ITAR / EAR | Defense manufacturing | US; mandatory for DoD contractors | CUI protection; supply chain flow-down; export control of technical data |
| NIS2 Directive | Essential and important entities | EU; mandatory for qualifying manufacturers | Risk management; incident reporting within 24h; supply chain security; board accountability |
EU Machinery Regulation and Cyber Resilience Act
The EU Machinery Regulation (2023/1230), replacing the Machinery Directive 2006/42/EC, introduces explicit cybersecurity requirements for machinery with digital elements for the first time. Starting January 2027, machinery placed on the EU market must ensure that its digital components do not compromise safety when connected to external networks or when subject to malicious interference. This is a fundamental shift: previously, CE marking for machinery only considered functional safety, not cybersecurity. For manufacturers who export to the EU, this means product cybersecurity must be integrated into the design phase, not bolted on after development.
The Cyber Resilience Act (CRA) extends further, requiring manufacturers of products with digital elements to maintain vulnerability handling processes, provide security updates for the expected product lifetime, and include an SBOM (Software Bill of Materials) with each product. For industrial IoT devices, PLCs with embedded web servers, and smart sensors, the CRA means ongoing cybersecurity obligations that persist long after the product is sold. The penalty regime is significant: up to 15 million EUR or 2.5% of global annual turnover for non-compliance.
A mid-sized US precision machining company manufacturing parts for a Tier 1 defense contractor must simultaneously comply with CMMC 2.0 Level 2 (110 NIST SP 800-171 controls for CUI protection), ITAR (International Traffic in Arms Regulations for technical data control), NIST Manufacturing Profile (for OT-specific risk management), and their prime contractor's flow-down requirements. The company's OT network processes CUI — technical drawings and CNC programs for controlled items. This means CMMC scoping includes the OT environment, requiring access controls, audit logging, and encryption on systems that were never designed for these capabilities. The compliance cost exceeded $400,000 — a significant burden for a company with 150 employees.
Building a Multi-Standard Compliance Matrix
The practical approach to multi-standard compliance is a unified control matrix that maps each control you implement to every standard it satisfies. Start with IEC 62443 as the base framework — it is the most OT-specific and the most detailed. Map each 62443 requirement to corresponding controls in NIST CSF, NIST SP 800-171 (if defense), NIS2 (if EU), and any sector-specific requirements. A single control — for example, network segmentation between IT and OT — satisfies requirements in IEC 62443 (zone/conduit), NIST CSF (PR.AC-5), NIS2 (Article 21(2)(a)), and CMMC (SC.L2-3.13.1). By implementing the control once and mapping it to all applicable standards, you avoid duplicative compliance work and demonstrate a coherent security program to auditors, regulators, and customers.
Insurance Requirements for OT Environments
Cyber insurance underwriters have become increasingly sophisticated about OT risk. Renewal questionnaires now routinely ask about OT/IT segmentation, OT-specific monitoring capabilities, SIS isolation, and backup/recovery procedures for control system configurations. Failure to demonstrate these controls results in coverage exclusions or premium increases of 200–400% for manufacturers with OT environments. The insurance questionnaire is a useful forcing function for security investment — when the CFO sees a $500,000 premium increase tied to the absence of OT monitoring, the budget conversation changes. Build your compliance matrix to include insurance questionnaire requirements alongside regulatory standards.
Do not build separate compliance programs for each standard. Build one security program based on IEC 62443, then create mapping layers that translate your controls into the language and numbering of each standard or regulation you must address. This approach reduces audit fatigue, ensures consistent implementation, and makes it possible for a mid-sized manufacturer to manage compliance across IEC 62443, NIST, NIS2, and CRA without tripling the compliance team. The compliance matrix is a living document — assign ownership, review quarterly, and update it as regulations evolve.
This lesson summarises the documents below. Where it matters — a filing, an audit response, a board paper — read the source rather than the summary. Every link was checked on 31 August 2026.
- MITRE ATT&CK
- CIS Critical Security Controls
- ISO/IEC 27001:2022 — Information security management systems (publisher blocks automated link checks; cited by name)