Risk Assessment Across the IT/OT Boundary
Build a structured risk register for IT/OT environments using threat modeling, consequence analysis, and the ISA/IEC 62443 risk framework.
Why Risk Assessment Comes First
Every control you deploy has a cost: hardware, licensing, operational overhead, and engineer time. Risk assessment answers the question of which threats deserve those costs — and in what order. Without it, you end up with a perfectly hardened historian and an unmonitored PLC exposed to the corporate VLAN.
ISA/IEC 62443-3-2 provides the formal methodology. This lesson distills it into a working process you can apply to a real site.
The Risk Assessment Process
62443-3-2 defines a structured, repeatable process:
1. Partition the system into zones and conduits
2. Perform an initial cybersecurity risk assessment per zone
3. Assign target security levels (SL-T)
4. Document required compensating controls
5. Reassess after control implementation (SL-A vs SL-T)
6. Establish an ongoing review cadence
Steps 1 and 3 were covered in Lesson 2. This lesson focuses on steps 2 and 4: how to quantify risk and translate it into control requirements.
Risk = Consequence × Likelihood
The 62443 framework defines risk as the product of consequence (what happens if the threat succeeds) and likelihood (how probable is successful exploitation). Both are scored on a 1–5 scale.
Consequence Categories
62443-3-2 uses these consequence dimensions:
| Dimension | Examples |
|---|---|
| Safety | Injury, death, mass casualty |
| Environment | Spill, release, regulatory violation |
| Financial | Downtime cost, remediation cost, fines |
| Reputational | Brand damage, public trust |
Score each dimension 1 (negligible) through 5 (catastrophic), then use the highest single dimension score as the zone’s consequence rating.
Likelihood Factors
Likelihood depends on:
- Threat actor capability — script kiddie vs. nation-state
- Attack surface — is the zone internet-accessible? is there a known exploit?
- Existing controls — does a firewall already block the attack vector?
Score 1 (very unlikely) through 5 (near-certain given current exposure).
Risk Matrix
| C1 | C2 | C3 | C4 | C5 | |
|---|---|---|---|---|---|
| L5 | Medium | High | High | Critical | Critical |
| L4 | Low | Medium | High | High | Critical |
| L3 | Low | Low | Medium | High | High |
| L2 | Low | Low | Low | Medium | High |
| L1 | Low | Low | Low | Low | Medium |
Use this matrix to prioritize remediation: Critical and High findings go in the first sprint; Medium and Low are scheduled.
Building the Risk Register
A risk register is a living document — not a one-time audit artifact. Each row is a specific threat against a specific zone:
Risk Register — IT/OT Security Assessment
──────────────────────────────────────────────────────────────────────────────
ID | Zone | Threat | C | L | Risk | Control
─────|───────────────|─────────────────────────────────|───|───|────────|──────────────────────────────
R-01 | Field Control | Unauthenticated Modbus command | 4 | 4 | High | Zeek + Suricata whitelist rules
R-02 | OT DMZ | Unencrypted historian sync | 3 | 3 | Medium | WireGuard tunnel on port 8089
R-03 | IT Enterprise | Hardcoded API key in pipeline | 3 | 4 | High | HashiCorp Vault dynamic secrets
R-04 | Field Control | Firmware modification via USB | 5 | 2 | High | USB port disable + asset inventory
R-05 | OT DMZ | Jump server credential theft | 4 | 3 | High | MFA + session recording on jump server
R-06 | Field Control | Safety interlock command bypass | 5 | 2 | High | Inline PLC command validator (M5)
R-07 | IT Enterprise | Unsigned build artifact deployed | 3 | 3 | Medium | Sigstore/cosign pipeline signing (M3)
Start with this YAML structure — it’s machine-readable and can feed a dashboard or ticketing system:
# risk-register.yaml
risks:
- id: R-01
zone: Field Control
conduit: OT-DMZ to Field Control
threat: Unauthenticated Modbus TCP command injection
threat_actor: Intentional, moderate capability (SL 2)
consequence: 4 # serious safety consequence if speed setpoint manipulated
likelihood: 4 # Modbus has no native authentication; directly reachable
risk: High
target_sl: 2
controls:
- description: Zeek script to whitelist valid Modbus function codes per device
status: planned
module: 4
- description: Suricata rule to alert on out-of-range register writes
status: planned
module: 4
residual_likelihood: 2 # after controls applied
residual_risk: Medium
owner: OT Security Team
review_date: 2026-12-01Iterate this file as you implement each module. The residual_likelihood field should decrease as controls go live — if it doesn’t, your control isn’t working.
Threat Scenario Deep-Dive: Modbus Command Injection
Walk through R-01 end-to-end to understand how abstract risk scores translate to real consequences.
The Asset
A variable frequency drive (VFD) controls motor speed at an industrial pump station. The PLC sends Modbus TCP write commands to register 40001 (speed setpoint) over the OT network.
The Threat
An attacker who has lateral-moved from a compromised engineering workstation sends a crafted Modbus TCP write command: register 40001, value 0xFFFF (65535 RPM — far beyond the motor’s rated 1800 RPM).
The PLC has no authentication on Modbus. It accepts the command and drives the motor at a destructive speed. The safety interlock (the SIS) should catch this — but if the attacker also disables the high-speed switch sensor (as Stuxnet disabled centrifuge sensors), the SIS never fires.
The Consequence
Motor bearing failure, potential shaft separation, and — in a water treatment context — loss of service to a population center.
The Controls
| Control | How It Helps | Module |
|---|---|---|
| Network segmentation | Attacker can’t reach the PLC without crossing an inspected boundary | M2 |
| Zeek whitelist | Logs any Modbus write to 40001; alerts on unexpected source IPs | M4 |
| Suricata rule | Blocks Modbus write if value > safe_max (1800) inline | M4 |
| Inline PLC validator | Last-mile check: rejects the command before the PLC receives it | M5 |
Defense in depth means the attacker must defeat all four controls, not just one.
Consequence Analysis for Safety Systems
When the consequence dimension is safety, standard financial risk math is insufficient. A 1-in-1000-year event that kills 50 people is not acceptably managed by buying insurance.
ALARP (As Low As Reasonably Practicable) is the engineering standard: risk must be reduced until the cost of further reduction is grossly disproportionate to the benefit gained. For safety-critical OT systems, this translates to:
- Every identified High risk gets a control, regardless of cost, until it is at least Medium
- Critical risks get controls until they reach Low or until the site can demonstrate that further controls are technically infeasible
- Residual risk acceptance must be documented and signed by the plant manager (not just the security team)
Safety Integrity Level (SIL) from IEC 61511 and Security Level (SL) from IEC 62443 are different standards with different numbering that happen to overlap. A SIL 3 safety function is not the same as an SL 3 security requirement, and satisfying one does not satisfy the other.
In fact, a system can have a SIL 3 safety function and an SL 1 security posture — meaning its safety interlock works correctly under normal conditions but can be defeated by a motivated attacker. Triton/TRISIS exploited exactly this gap: the SIL-certified safety system had no cybersecurity controls at all.
Track both independently in your risk register.
The Review Cadence
A risk register that isn’t updated is an audit document, not a security tool. Schedule reviews triggered by:
- New asset added to a zone (e.g., a new PLC model or historian upgrade)
- New threat intelligence (e.g., a published exploit for Modbus-enabled devices)
- Control change — a control modified, disabled, or bypassed
- Incident — any security event, even minor ones
- Annual minimum — regardless of changes, per 62443-2-1 requirements
A zone assessment gives consequence = 5 (catastrophic safety risk) and likelihood = 2 (unlikely — existing firewall provides some protection). What is the risk level, and what does ALARP require?
Reveal Answer & Explanation
Using the risk matrix: High (C5, L2 = High).
ALARP requires that every High finding receive controls until the residual risk is at least Medium. The existing firewall that reduces likelihood to 2 is already counted — the residual risk after controls must be reduced further.
Acceptable next controls might include: adding DPI to the firewall (reducing likelihood to 1, dropping the risk to Medium), or reducing consequence by installing a hardware safety interlock that is physically independent of the compromised zone.
Once you’ve applied controls and reassessed, the residual risk must be formally accepted by the plant manager in writing. “We knew and chose not to act” is not ALARP — it’s negligence.
What’s Next
You have a structured risk register and understand how to translate threat scenarios into security level requirements.
In Module 2, we begin implementing controls: deploying WireGuard to secure the authenticated network transport across IT/OT conduits — addressing the network-layer gaps surfaced in your risk register.