Service

SCADA and ICS Security

Industrial control system security designed around the constraint that the plant cannot stop and the controller cannot be patched.

Home Services SCADA & ICS security

  • Purdue zones
  • Passive only
  • Modbus / DNP3 / PROFINET
  • Safety-first

In plain terms

The computers that physically run pumps, production lines and power equipment were designed before networks were hostile, and most cannot be patched or replaced for a decade. We protect them by controlling the network around them and watching for anything abnormal — never by installing software on them or scanning them.

SCADA security fails when IT practice is applied unchanged. An active scan can knock over a PLC. An agent rollout can void a vendor warranty. The architecture has to assume none of that is available.

Zone and conduit architecture

OT and SCADA zone and conduit modelEnterprise IT, a DMZ data broker, a supervisory layer and three independently segmented process cells, with inspected conduits between zones and passive monitoring inside them.Enterprise ITcorporate email, ERP, general user accessinspected conduitDMZ / data brokerhistorian replica, jump host, one-way where possibleinspected conduitSupervisory — HMI, SCADA serversCELL-LEVEL SEGMENTATION — PASSIVE MONITORING ONLYCell APLCs, drivesno agent permittedCell Bsafety instrumentednever auto-isolatedCell Cfield devices, RTUs
Where the device cannot be secured, the network around it is. Safety-instrumented cells are marked never-auto-isolate.

The design principle is that traffic crossing a zone boundary passes through an inspected conduit, and traffic inside a cell is observed but never interfered with. Where the process allows it, the boundary between the DMZ and the supervisory layer is made one-way — a data diode or an equivalent — so no inbound path exists at all.

How we deploy without touching the process

  1. Passive captureSPAN or TAP only. No active scanning, ever. An unexpected probe against a safety instrumented system is an incident in its own right.
  2. Protocol decodeModbus, DNP3, BACnet, PROFINET, EtherNet/IP and vendor-proprietary traffic decoded so a setpoint change reads as a setpoint change.
  3. BaselineIndustrial traffic is far more regular than IT traffic, which makes anomaly detection genuinely effective here — but only after a full production cycle is observed.
  4. Boundary enforcementConduits implemented and tested during a planned outage, with the operations team present and a rehearsed rollback.
  5. ExclusionsSafety systems registered as never-auto-isolate before any automation is enabled.

SCADA security consulting and assessment

Engagements usually start as one of two things. A SCADA security assessment is a fixed-scope review: passive asset discovery, conduit and boundary analysis, a mapped set of SCADA security vulnerabilities with exploitability weighted for an OT context, and a prioritised remediation plan that respects your outage calendar. SCADA security consulting is the longer engagement that follows — architecture, implementation and the operating model.

The SCADA security controls we assess against

  • Zone and conduit design — IEC 62443 zones, with every conduit justified and inspected
  • ICS/SCADA network security — segmentation between supervisory and process cells
  • SCADA HMI security — workstation hardening, application allow-listing, and the removable media path that is still the most common entry point
  • Authentication and least privilege — including the shared operator accounts that every plant has and few document
  • Remote and vendor access — usually the widest and least monitored path in
  • Backup and recovery — controller configurations and logic, restore-tested

SCADA security standards we map to

IEC 62443 for the architecture, NIST SP 800-82 for the control set, and NERC CIP where you are a registered entity. We map once to the underlying mechanism and derive each framework view from it, rather than running three parallel compliance exercises.

The rule we do not break

Nothing is deployed against a control system without operations in the room and a tested rollback. A security control that trips the process has caused exactly the outcome it was bought to prevent.

Unidirectional gateways

Unidirectional gateway between process and enterprise networksA primary historian writes through a transmit-only optical interface to a replica on the enterprise side; no return path exists at the physical layer.PROCESS SIDE — TRUSTEDHistorian (primary)writes continuouslyTransmit-only interfaceoptical TX, no RX pathone wayphysics,not configENTERPRISE SIDE — UNTRUSTEDHistorian replicaread-only mirrorBusiness reporting, dashboardsthere is no return path for an attacker to use, because there is no return path at allthe cost: no remote commands, no remote patching, no pull-based polling — everything inbound must be physically escorted
A firewall rule can be misconfigured. A diode cannot — there is no receiver on the process side to misconfigure.

Where the process tolerates it, the strongest boundary is not a firewall rule but the absence of a receiver. A data diode permits historian data out to enterprise reporting and makes an inbound path physically impossible. The cost is real — no remote commands, no remote patching, no pull-based polling — and we scope that honestly before recommending one.

Choosing a boundary control

ControlInbound possibleCostUse when
Firewall rulesYes, by configurationLowInbound access is genuinely required
Broker in a DMZOnly to the brokerMediumBidirectional need, but no direct path acceptable
Data diodeNo — physicallyHighOutbound reporting only; highest-consequence processes

The SCADA threat model

SCADA attack paths and why they succeedFive initial access routes — vendor remote access, the IT network, removable media, insiders and internet exposure — converge on an engineering workstation, from which unauthenticated legacy protocols permit command spoofing with physical consequence.five ways in — only one of them is the internetVendor remote accesssupply chainIT networkIT/OT convergenceRemovable mediaUSB to workstationInsiderdeliberate or errorInternet-exposeddevice or HMIEngineering workstation / HMILegacy protocol, no authenticationModbus and DNP3 travel in clear textCommand spoofing — the physical consequencealter setpoints · force coils · disable alarms · override safetyWHAT MAKES IT WORKOnce an attacker reaches a workstation the protocols offer no resistance — they were designed for a network assumed to be physically safe.
Note how few of these involve the internet. Segmentation and vendor access control cover more real attack paths than perimeter hardening does.

SCADA security is a different category of risk from ordinary IT security because it bridges the digital and the physical. An IT breach costs you data. A successful SCADA attack can stop a production line, contaminate a water supply, or open a breaker on a grid. Legacy SCADA was designed for isolated environments, long before the network was hostile, so encryption and authentication were simply not part of the design.

Threat types

ThreatHow it works herePhysical consequence
Industrial malware Purpose-built code targeting control systems. Stuxnet altered centrifuge logic while reporting normal readings to operators; Industroyer spoke grid protocols natively to open breakers. Equipment damage, process manipulation invisible to the control room
Ransomware Usually not OT-aware. It encrypts the engineering workstations and HMIs, and production stops because operators are blind — not because controllers were touched. Production halt, extortion
Command spoofing Legacy protocols such as Modbus and DNP3 transmit in clear text with no authentication. Anyone who can reach a device can issue a valid-looking command. Altered setpoints, forced valve states, suppressed alarms
Denial of service Flooding the network blocks telemetry between field devices and the control room. Operators blinded during a live physical emergency
Supply chain Vendors need remote access to maintain equipment. Compromise the vendor and inherit their path into every customer. Trusted access, no perimeter to cross

Threat actors, and what each one wants

ActorMotivationWhat it means for your defenses
Nation-state (APT)Strategic sabotage, pre-positioning for future conflict Patient, well-resourced, will spend months quiet. Detection matters more than prevention.
CybercriminalsExtortion payouts Opportunistic. Basic segmentation and backups defeat most of them.
CyberterroristsPublic disruption and panic Target visibility over value — water and power attract disproportionate attention.
InsidersGrievance, or simple error Bypass the perimeter entirely. An infected USB drive in a control workstation is the classic unintentional case, and it is far more common than sabotage.

The underlying vulnerabilities

  • IT/OT convergence. Connecting plant systems to the corporate network for reporting exposes equipment that was previously unreachable. This is the single largest structural change in industrial risk over the past decade, and it was driven by business reporting needs, not by anyone deciding to accept the risk.
  • Unencrypted legacy protocols. Many controllers lack the processing power for modern cryptography, so traffic can be read and forged. Encryption is not coming to these devices — the network around them has to carry the burden.
  • Weak credential management. Hardcoded manufacturer passwords, shared operator accounts and no multi-factor authentication. Default credentials for most industrial equipment are published in the vendor's own manual.
  • Removable media. Where the network is genuinely segmented, USB becomes the transfer mechanism — and therefore the infection path. Segmentation without a media control policy relocates the problem rather than solving it.

What this changes about defense

Four of the five access paths in the diagram above do not involve the internet. A program that invests entirely in perimeter hardening has addressed one of them. Vendor access control, IT/OT boundary enforcement, removable media policy and insider-capable monitoring cover the other four — and they are consistently the cheaper work.

SCADA cyber security terminology

SCADA security, ICS/SCADA security and OT security are used more or less interchangeably in the market, and the distinctions matter mostly when writing a scope of work.

  • SCADA refers specifically to supervisory control and data acquisition — the supervisory layer that monitors and controls a process.
  • ICS (industrial control systems) is the broader family, including DCS, PLCs and safety instrumented systems. Most "SCADA cyber security" engagements are really ICS SCADA cyber security in scope.
  • OT is broader again, covering building systems and any technology that interacts with the physical world. See OT security solutions.

Our SCADA security best practices follow IEC 62443 for architecture and NIST SP 800-82 for the control set, and a cyber security risk assessment for SCADA and DCS networks is scoped the same way regardless of which label the client uses. Security for SCADA systems is ultimately a segmentation and monitoring problem, because the endpoints themselves usually cannot be secured directly.

The recurring SCADA security issues we find are consistent across sectors: expired firewall exceptions left from a commissioning task, shared operator credentials, unmonitored vendor remote access, and controller logic with no verified backup.

IEC 62443 zones in practice

LevelContentsTypical conduit
4–5 EnterpriseERP, email, general ITTo DMZ only, authenticated
3.5 DMZHistorian replica, patch server, jump hostBroker; ideally push-only from below
3 OperationsSCADA servers, engineering workstationsInspected, protocol-aware
2 SupervisoryHMI, alarm systemsCell-restricted
1 ControlPLCs, RTUs, DCS controllersWithin cell only
0 ProcessSensors, actuatorsPhysical
SISSafety instrumented systemsSeparate. Never auto-isolated. Frequently air-gapped.

Worked example — discrete manufacturer, ransomware in the plant

An intrusion that began on corporate reached the plant supervisory network and encrypted two HMI workstations. Production stopped for 31 hours. We were engaged for recovery and then for the rebuild.

  1. How it reached the plantA single firewall rule, added four years earlier for a vendor commissioning task and never removed, permitted RDP from a corporate subnet to the supervisory VLAN. This is the most common finding in OT intrusions and it is almost always an expired exception nobody reviewed.
  2. Why it stopped where it didThe PLCs were untouched — not because they were protected, but because the ransomware had no Modbus capability. That is luck, and we said so rather than crediting the architecture.
  3. RecoveryHMIs rebuilt from images. Controller logic was intact but there was no verified backup of it — the team's confidence that "the PLCs have the logic" held only because nothing had corrupted them.
  4. Rebuild — conduitsEvery IT/OT rule re-derived from observed traffic. 34 existing rules reduced to 6, of which 2 were push-only from the OT side.
  5. Rebuild — vendor accessReplaced with a brokered, time-boxed, recorded session requiring operator approval to open. Vendors disliked it; it was non-negotiable.
  6. Rebuild — controller backupAutomated extraction and versioning of PLC logic with restore testing each quarter. They had operated for eleven years without this.
  7. Rebuild — monitoringPassive sensors per cell with baselines from a full production cycle including changeovers.

The rule-review finding, generalized

Every OT estate we assess has firewall exceptions added for a commissioning task and never removed. We now treat rule expiry as a control in its own right: every IT/OT rule carries an owner and a review date, and rules without both are removed at the next window.

Common questions

What is SCADA security?

SCADA security protects the supervisory control and data acquisition systems that operate physical processes in utilities, manufacturing and infrastructure. It differs from IT security because availability and safety outrank confidentiality, devices often cannot be patched or scanned, and a failed control can cause physical consequence rather than data loss.

What are the main SCADA security vulnerabilities?

Most commonly: flat networks with no segmentation between business and process systems, unauthenticated industrial protocols that were never designed for a hostile network, unpatchable controllers running long past vendor support, shared operator credentials, and remote vendor access paths that are neither monitored nor time-bounded.

Scope your endpoint and network architecture

Bring us the estate you actually have — mixed OS fleets, ageing routers, unmanaged IoT — and we will tell you what can be unified under one agent and what needs custom work.

Request a FREE consultation