Service
Industrial control system security designed around the constraint that the plant cannot stop and the controller cannot be patched.
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.
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.
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.
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.
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.
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.
| Control | Inbound possible | Cost | Use when |
|---|---|---|---|
| Firewall rules | Yes, by configuration | Low | Inbound access is genuinely required |
| Broker in a DMZ | Only to the broker | Medium | Bidirectional need, but no direct path acceptable |
| Data diode | No — physically | High | Outbound reporting only; highest-consequence processes |
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 | How it works here | Physical 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 |
| Actor | Motivation | What 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. |
| Cybercriminals | Extortion payouts | Opportunistic. Basic segmentation and backups defeat most of them. |
| Cyberterrorists | Public disruption and panic | Target visibility over value — water and power attract disproportionate attention. |
| Insiders | Grievance, 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. |
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 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.
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.
| Level | Contents | Typical conduit |
|---|---|---|
| 4–5 Enterprise | ERP, email, general IT | To DMZ only, authenticated |
| 3.5 DMZ | Historian replica, patch server, jump host | Broker; ideally push-only from below |
| 3 Operations | SCADA servers, engineering workstations | Inspected, protocol-aware |
| 2 Supervisory | HMI, alarm systems | Cell-restricted |
| 1 Control | PLCs, RTUs, DCS controllers | Within cell only |
| 0 Process | Sensors, actuators | Physical |
| SIS | Safety instrumented systems | Separate. Never auto-isolated. Frequently air-gapped. |
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.
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.
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.
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.
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.