Skip to content
Tutorials Center
My learning

Topics

Lesson 66 of 80 ยท Privacy, Cloud and Secure Design

Critical Infrastructure and ICS Security

Learn why power grids, water, transport and factories need special security: ICS and SCADA components, the Purdue model, typical threats, and practical defense and monitoring.

Some computer systems do not just store information; they move physical things. A control system opens a valve in a water plant, trips a breaker on the electrical grid, switches a railway signal or regulates the temperature of a chemical reactor. When such systems fail or are manipulated, the result is not just a data breach but blackouts, polluted water, stopped trains or injured people. The facilities and services that societies depend on (energy, water, transport, healthcare, communications, food, finance, manufacturing) are called critical infrastructure.

Security for these environments is called ICS security (industrial control systems), or OT security (operational technology), as opposed to ordinary IT security. The goals and constraints differ, so the usual office-network advice, such as "patch everything at once" or "scan the network aggressively", can itself cause harm. This lesson explains the components, the differences from IT, the main threats and historical lessons, the Purdue reference model, and defensive practices. It is entirely about defence; no attack techniques are given. It builds on Firewalls, Segmentation and Zero Trust and Threats, Vulnerabilities and Risk.

What is in an industrial control system?

  • Sensors and actuators: devices that measure (temperature, pressure, flow, voltage) and act (valves, pumps, motors, breakers).
  • PLCs (programmable logic controllers) and RTUs (remote terminal units): ruggedized computers that read sensors and drive actuators according to a stored control program, in a continuous loop, often in milliseconds.
  • DCS (distributed control system): control for a single large site such as a refinery or power plant.
  • SCADA (supervisory control and data acquisition): software and networks that gather data from many, often geographically spread, sites (pipelines, power grids, water networks), display it, and send commands.
  • HMI (human-machine interface): the screens operators use to see and control the process.
  • Historian: a database recording process values over time.
  • Safety instrumented systems (SIS): independent systems that bring a process to a safe state when limits are exceeded.
  • Engineering workstations: where control programs are written and loaded into controllers.

Many industrial protocols (such as Modbus and DNP3, and older variants of other fieldbus protocols) were designed decades ago for isolated serial networks, so they typically have no authentication or encryption: any device that can send a valid message may be obeyed. Newer versions and gateways add security, but old equipment stays in service for 20 to 30 years.

How OT differs from IT

IT security OT / ICS security
Top priority Confidentiality, then integrity, then availability Safety and availability first, then integrity; confidentiality is often last
Failure impact Data loss, downtime, fraud Physical harm, environmental damage, outages of essential services
Lifespan 3 to 5 years 15 to 30 years
Patching Frequent, can reboot Rare and scheduled around shutdowns; vendor must certify; some devices cannot be patched
Timing Delays tolerated Real-time and deterministic; added latency or scanning can disrupt the process
Operating systems Current, standard Legacy or embedded systems, sometimes unsupported
Typical tools Active scanning, EDR on everything Passive monitoring; agents may not be allowed
People Security team Engineers and operators who know the process; security must work with them

Note the effect on the CIA triad: the order of priorities can reverse, and integrity includes the integrity of process values (is that pressure reading real?) and commands.

Threats and lessons from history

Critical infrastructure attracts nation-states, criminal groups and insiders, and it also suffers from accidents and misconfiguration. Typical threat types:

  • Ransomware on IT that spreads to or forces shutdown of OT for safety or because the business systems that schedule production are down.
  • Targeted intrusions that study the process and manipulate it, or prepare for later disruption (The Attack Lifecycle and MITRE ATT&CK includes an ICS matrix).
  • Compromised remote access: vendor and operator VPNs, remote desktop, and shared credentials.
  • Supply-chain compromise: tainted firmware, engineering software or a maintenance provider (Dependency and Supply Chain Security).
  • Insecure protocols and flat networks: a single foothold can reach controllers.
  • Insiders and mistakes: unauthorized changes, USB drives, laptops that move between networks.
  • Physical threats and DDoS, and for geographically spread systems, weak or unprotected field sites.
  • Environmental and supplier failures that affect resilience.

Landmark incidents shaped the field. Stuxnet (discovered 2010) showed that malware can target specific industrial controllers and cause physical damage; attacks on the Ukrainian power grid (2015 and 2016) caused real outages using stolen credentials, remote access and operator-interface tools; the Triton/Trisis malware (2017) targeted safety systems at a petrochemical site; and the Colonial Pipeline ransomware incident (2021) caused a fuel pipeline to shut down proactively because business systems were disabled, even though the control system itself was reportedly not compromised. The shared lesson is that IT compromise can cause OT outages, remote access is a favourite route, and safety systems themselves need protection.

Sectors and their particular challenges

  • Energy and electric grid: generation, transmission and distribution depend on real-time balance; digital substations and growing numbers of distributed energy resources (solar, batteries, smart meters) widen the attack surface.
  • Water and wastewater: many small utilities with limited budgets and staff, remote sites and old controllers.
  • Transport: railway signalling, air-traffic systems, ports, traffic control and increasingly connected vehicles.
  • Manufacturing and process industries: robots, production lines, chemical and food processing, increasingly linked to business systems and the cloud (Industry 4.0 and industrial IoT).
  • Healthcare: medical devices that cannot easily be patched and must keep working.
  • Building systems: HVAC, access control and elevators, often managed by third parties.

The Purdue model and segmentation

A widely used reference architecture, the Purdue model (and the ISA/IEC 62443 zones-and-conduits concept), divides an industrial company's networks into levels:

Level Contains Role
0 Sensors, actuators, drives The physical process
1 PLCs, RTUs, safety controllers Basic control
2 HMIs, SCADA servers, engineering stations Supervisory control for one area
3 Site operations: historians, patch and application servers Manufacturing operations
3.5 Industrial DMZ Controlled exchange between IT and OT
4-5 Business and enterprise networks, internet Corporate IT

The key idea is segmentation with controlled conduits: traffic should flow only between adjacent levels, through firewalls and the DMZ, and never directly from the internet or office network to a controller. Modern architectures blur the levels with cloud analytics and remote access, but the principle of zones and conduits (group assets by function and risk, and control every path between zones) remains the core design tool (IEC 62443 is the main standards family).

flowchart TB
    L5["Levels 4-5: Enterprise IT and internet"] --> DMZ["Level 3.5: Industrial DMZ (firewalls, jump hosts, data diodes)"]
    DMZ --> L3["Level 3: Site operations (historian, patch server)"]
    L3 --> L2["Level 2: HMI and SCADA servers"]
    L2 --> L1["Level 1: PLCs and RTUs"]
    L1 --> L0["Level 0: Sensors and actuators"]

Each arrow is a controlled conduit; direct jumps (for example from the office network to a PLC) are what segmentation forbids.

The checker below encodes a simple policy of which level-to-level flows are allowed and tests proposed firewall rules against it:

Python Check flows against the Purdue zones

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

Defensive practices

Know what you have. Build an asset inventory of controllers, firmware versions, network paths and owners. You cannot protect devices you do not know about.

Segment and control paths. Separate IT from OT with an industrial DMZ; separate zones inside OT; allow only the protocols and endpoints needed; block direct internet access from control networks. For the highest-risk one-way flows (such as sending historian data out) a data diode enforces direction physically.

Secure remote access. Use a single, monitored path with multi-factor authentication, individual accounts (no shared passwords), time-limited vendor sessions, jump hosts and session recording (Multi-Factor Authentication and Passkeys).

Harden and patch with care. Remove unused services, change default passwords, restrict engineering workstations, control USB and removable media, and apply vendor-certified patches during planned outages. Where patching is not possible, use compensating controls: network isolation, virtual patching and monitoring (Vulnerability Management and Patching).

Monitor passively. Network sensors that understand industrial protocols (listening on mirror ports, not actively probing) learn what normal looks like and alert on new devices, new connections, unexpected commands (such as a program download to a controller) and values outside safe ranges. Combine with logs from firewalls and servers (Network Traffic Analysis and Packet Capture, Logging, Log Analysis and SIEM).

Protect safety systems. Keep safety controllers independent, restrict who can reach and reprogram them, and keep key switches in the "run" position except during authorized changes.

Back up and recover. Keep offline copies of controller programs and configurations and known-good firmware; test restoration and manual operation modes (Backups, Resilience and Business Continuity).

Prepare incident response. Plans must include engineers and operators, safe-shutdown criteria and communication with regulators and utilities (The Incident Response Process). Practise with tabletop exercises that include "what if the screens lie?".

Manage people and suppliers. Train operators to recognize social engineering, require security from vendors, and review contracts and remote-support access (Phishing and Social Engineering).

Monitoring process values

One useful OT detection is to compare readings against physical limits and plausible rates of change. A tank cannot jump 40 percent in one second; a command to run a pump when the tank is full is suspicious. The simulation below watches synthetic readings and commands for violations of both kinds of rule:

Python Detect implausible process values and unsafe commands

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

Rules like these are cheap and independent of any malware signature, because physics does not lie (provided redundant sensors exist). Operators and security teams build them together.

Standards, laws and frameworks

  • IEC 62443 / ISA-99: the main industrial security standards family: zones and conduits, security levels, secure development and component requirements.
  • NIST SP 800-82: guide to OT security; NIST Cybersecurity Framework adapted to ICS (Compliance and Security Frameworks).
  • Sector rules: for example NERC CIP for the North American bulk electric system, and the EU NIS2 directive covering essential and important entities.
  • MITRE ATT&CK for ICS: a knowledge base of adversary behaviour against control systems.
  • Many countries require incident reporting for critical infrastructure operators (Cyber Law, Ethics and Security Policy).

Check your understanding

Quick check ยท 4 questions

of 4 answered You scored / 4 ยท perfect!

  1. 1. Which priority order is typical for OT systems?

    Industrial processes must keep running safely, so safety and availability usually outweigh confidentiality, the reverse of many IT systems.
  2. 2. Why is aggressive active scanning with standard IT tools risky on a control network?

    Many controllers run old, fragile network stacks and tight timing loops. Passive monitoring is preferred, and any active testing is planned with engineers.
  3. 3. What is the purpose of the industrial DMZ (level 3.5) in the Purdue model?

    Services such as jump hosts, patch servers and data replication live in the DMZ, so IT systems do not talk straight to controllers.
  4. 4. The Colonial Pipeline incident (2021) mainly taught which lesson?

    Business systems needed for billing and scheduling were affected, and the operator halted operations as a precaution, showing the dependency between IT and OT.

Summary

  • Critical infrastructure combines IT with physical processes, so failures can harm people and the environment.
  • ICS components include sensors, actuators, PLCs/RTUs, HMIs, SCADA servers, historians and safety systems; many protocols lack authentication.
  • OT puts safety and availability first, runs legacy equipment for decades, patches rarely and needs passive, engineer-approved security measures.
  • Threats include ransomware spill-over, remote-access abuse, supply-chain compromise and targeted manipulation; Stuxnet, the Ukrainian grid attacks, Triton and Colonial Pipeline are the key case studies.
  • The Purdue model and zones-and-conduits segment the network, with an industrial DMZ between IT and OT.
  • Defend with inventory, segmentation, secured remote access, careful patching, passive monitoring including process-value checks, protected safety systems, offline backups and joint IT-OT incident response.

Next: Cyber Law, Ethics and Security Policy.