OT Cybersecurity
Cloud and OT security - threats, architecture, and controls for industrial cloud deployments
Cloud threats to OT networks - cloud SCADA, IIoT gateways, shared responsibility, zero trust cloud-OT. Guide with checklist and scenarios.
Michał StępieńPublished Updated 14 min read

Contents
- How cloud enters OT environments
- Threat scenarios - from data manipulation to a path into the OT network
- The shared responsibility model for cloud-OT environments
- Reference architecture - securely connecting cloud to OT
- Zero Trust for cloud-OT environments
- Checklist: before deploying a cloud service in an OT environment
- Common mistakes when deploying cloud in OT
- Incidents from 2024-2025
- Cloud-OT risk assessment per IEC 62443
- Regulation and the supply chain
- Where we help
- Sources
Connecting an OT network to the cloud gives the operator data and remote access. It also creates a new path to the controllers and a new dependency on the cloud provider.
For decades OT networks operated in isolation, physically separated from the outside world. Today data from OT networks reaches the cloud through SCADA platforms, IIoT gateways, predictive maintenance services and remote monitoring of renewable energy installations. The boundary between the OT network and the cloud becomes porous as a result.
Security models from IT protect mainly the confidentiality of data. In OT process continuity matters most, so those models have to be adapted.
How cloud enters OT environments
Cloud services enter OT through four main paths. Each introduces a different risk profile.
| Path | Example deployment | What goes to the cloud | OT risk |
|---|---|---|---|
| Cloud SCADA and historian | Historian in the SCADA vendor’s cloud | Process data, alarms and trends go to the cloud. | An attacker can manipulate process data, and the historian can become unavailable. |
| IIoT gateways | IIoT gateway with a modem and a manufacturer portal | Sensor telemetry and device diagnostics go to the cloud. | An attacker who takes over the gateway from the cloud side can reach the OT network through it. |
| Predictive maintenance | Public cloud IoT service with an edge module | Vibration and temperature data and operating parameters go to the cloud. | False models lead to poor maintenance decisions. |
| Remote monitoring of renewables | Management platform for wind and PV farms | Generated power, inverter status and setpoints go to the cloud. | Compromising the cloud platform allows setpoints to be changed. |
WARNING
Before an IIoT gateway is connected to the cloud, default passwords must be changed and certificate authentication enabled. Device data collected in the cloud needs protection too: in February 2025 an unprotected database of a manufacturer of home IoT grow lights exposed 2.7 billion records, including Wi-Fi network names and passwords.
Threat scenarios - from data manipulation to a path into the OT network
Scenario 1: Manipulation of cloud-OT communication
An attacker intercepts or modifies data transmitted between local controllers and cloud applications. If a predictive model in the cloud receives falsified process data, it generates incorrect setpoints. In HVAC systems in smart buildings, such as those described in the smart building cybersecurity article, incorrect control leads to discomfort or installation damage.
Attack sequence: the attacker takes over the cloud API and sends modified setpoints. The controller applies the new parameters and raises no alarm, because the values are within range, although not optimal.
Scenario 2: Escalation from cloud to OT network
An IIoT gateway connects “up” to the cloud and “down” to the industrial network. Compromising an administrative account in the cloud can give the attacker access to the gateway management interface, and from there to the OT network.
Attack sequence: the attacker phishes the credentials of a cloud administrator account and logs in to the gateway management portal. The attacker then pushes a firmware update with a backdoor and gains access to the OT segment.
In a 2024 Claroty survey, 82% of 1,100 respondents reported at least one attack on cyber-physical systems that came through third-party supplier access.
Scenario 3: Collateral damage in a multi-tenant environment
When data from multiple customers sharing the same cloud infrastructure is not properly isolated, an attack on one customer’s resources can disrupt the others. In OT this means that taking over a neighbouring tenant in a cloud SCADA platform can open a path to another organisation’s process data.
Scenario 4: Loss of cloud SCADA availability
A DDoS attack on the cloud provider’s infrastructure or an outage of a cloud region cuts the operator off from monitoring and historical data. If an organisation has moved its historian exclusively to the cloud, without a local copy, it loses process visibility at the most critical moment.
The shared responsibility model for cloud-OT environments
The shared responsibility model known from IT requires adaptation to OT specifics. The closer to the physical process, the more the operator is responsible.
| Domain | Cloud provider | Integrator or OT vendor | Operator |
|---|---|---|---|
| Physical data centre infrastructure | Responsible | - | - |
| Hypervisor and compute layer | Responsible | - | - |
| IoT or cloud SCADA platform | Co-responsible | Co-responsible | Configures the platform |
| Identity and access (IAM) | Provides the infrastructure | - | Sets policies, enables MFA and reviews permissions |
| Data encryption in transit | Handles TLS termination | Provides device certificates | Enforces policies |
| Data encryption at rest | Encrypts by default with provider-managed keys | - | Decides whether to manage the keys itself (e.g. own KMS/HSM) |
| IIoT gateway configuration | - | Sets the default configuration | Hardens the configuration and installs patches |
| OT network segmentation | - | - | Full responsibility |
| Physical process safety | - | - | Full responsibility |
The operator, not the cloud provider, is always responsible for OT network segmentation and process safety.
A few numbers
| Value | What it means | Source |
|---|---|---|
| 21.1bn | At the end of 2025 the forecast was 21.1 billion connected IoT devices worldwide, against 18.5 billion at the end of 2024. | IoT Analytics, 2025 |
| 82% | In a survey, 82% of respondents reported at least one attack on cyber-physical systems within 12 months that came through third-party supplier access. | Claroty, 2024 |
| 38% | Of the most exposed OT and medical devices, 38% have no CVSS score of 9.0 or higher, so prioritising by CVSS alone misses them. | Claroty, 2024 |
| 47% | In the survey, 47% of respondents name OT skills gaps among the three largest obstacles to securing OT and IIoT. | PwC, Global Digital Trust Insights 2026 |
Reference architecture - securely connecting cloud to OT
A secure cloud-OT architecture rests on three principles: minimising data flow towards OT, unidirectional communication wherever possible, and isolating the IIoT gateway in a dedicated DMZ zone.
Architecture levels
Zone 1 - Process network (L0-L1): PLC and RTU controllers, sensors and actuators. The cloud has no direct access here. Communication only with the local historian or IIoT gateway via industrial protocols.
Zone 2 - Industrial DMZ (L3.5): IIoT gateway or data diode. This is the sole point of contact between the OT network and the outside world. The gateway collects data from the historian and transmits it unidirectionally to the cloud. A data diode eliminates the possibility of return communication.
Zone 3 - Corporate network (L4-L5): Connection to the cloud via an encrypted VPN tunnel. Identity management, access policies, SIEM/XDR.
Zone 4 - Cloud: Historian, dashboard, analytics and ML. Process data is read-only. No control signal should return from the cloud to OT without explicit authorisation and audit.
TIP
Where unidirectional communication is not feasible (for example remote setpoint updates), a separate channel for setpoint changes should be set up, with MFA, session recording and approval by the on-site operator. The ICS remote access guide describes the principles of secure remote access.
Zero Trust for cloud-OT environments
The perimeter model assumes that an isolated OT network is secure. A cloud connection removes that isolation. The Zero Trust model verifies every request, regardless of where it originates.
Zero Trust principles for connecting cloud to OT
| Principle | Application in cloud-OT |
|---|---|
| Verify explicitly | Every IIoT gateway connection to the cloud is authenticated with an X.509 certificate, not a password. |
| Least privilege | The gateway has read-only access to process data, and write-back requires escalation. |
| Assume compromise | Gateway-cloud traffic should be monitored for anomalies: unusual hours, data volume and new endpoints. |
| Microsegmentation | The IIoT gateway runs in a separate network zone, not in a shared VLAN with SCADA systems. |
| Continuous verification | Certificates are rotated, and the gateway is locked out automatically after a set number of failed authentications. |
Example controls for an IIoT gateway at successive SL levels (a proposal, not a requirement of the standard)
The target SL-T of a zone with an IIoT gateway follows from the risk assessment. The controls below are a proposal, not a requirement of the standard. The SR references come from IEC 62443-3-3 and the EDR reference from IEC 62443-4-2. Controls marked as good practice have no basis in the standard.
| Parameter | SL 1 | SL 2 | SL 3 |
|---|---|---|---|
| Gateway authentication | Password and IP allowlist (good practice) | X.509 certificate (SR 1.2) | Certificate with a hardware-protected private key (SR 1.9 RE 1) |
| Encryption | TLS 1.2 (good practice) | TLS 1.3 with mutual authentication (good practice) | TLS 1.3 with a data diode as a fallback (good practice) |
| Monitoring | Local logs | Centralised logging (SIEM) and continuous monitoring (SR 6.2) | Network traffic monitoring (NDR) and anomaly detection (good practice) |
| Firmware updates | Manual | Digitally signed, with authenticity and integrity verification (EDR 3.10 RE 1) | As at SL 2, with an additional integrity check (good practice) |
| Remote access | VPN with MFA (good practice) | Access granted for the duration of a task, with session recording and approval of access over untrusted networks (SR 1.13 RE 1) | No remote access, on-site only (good practice) |
Checklist: before deploying a cloud service in an OT environment
- Inventory: list all OT-cloud connections, including IoT devices connected without the security team’s knowledge, for example by maintenance
- Risk analysis: for each IIoT gateway and cloud SCADA instance, assess the consequences of compromise per IEC 62443 (availability, integrity, confidentiality)
- SLA and contracts: the cloud provider contract must define RPO and RTO for OT services, incident notification procedures and the right to audit
- Vendor certification: verify vendor certifications (ISO 27001, SOC 2 Type II) and conformity with IEC 62443-4-1 for the development process and IEC 62443-4-2 for components
- Private vs. public cloud: for critical processes, consider a private or hybrid cloud, which leaves the operator in control of data and process know-how
- Encryption: encrypt data in transit (TLS 1.3) and at rest (AES-256), and let the operator, not the provider, manage the keys
- Segmentation: the IIoT gateway must run in a dedicated DMZ zone, not in a shared network with SCADA, because compromising a gateway on a network with controllers gives access to every device in the segment; more in the OT segmentation guide
- Unidirectional communication: where possible, deploy a data diode or unidirectional gateway
- Monitoring: deploy network traffic monitoring (NDR) with OT protocol decoding at the gateway-process network boundary
- Contingency plan: describe operation when the cloud service is unavailable; local control must be self-sufficient
- Testing: run regular penetration tests covering the cloud-gateway-OT path, not just the IT network
Common mistakes when deploying cloud in OT
The following mistakes recur in industrial environments and increase the risk of compromise.
Mistake 1: IIoT gateway in the process network
An IIoT gateway is connected directly to the network where PLC controllers operate, without segmentation. Compromising the gateway then gives the attacker access to every device in the segment. The solution is a dedicated DMZ zone with a firewall at the boundary, described in the OT network segmentation article.
Mistake 2: Default credentials on gateways
A gateway with a default administrator password and a cloud connection gives access to the OT network to anyone who knows that password. The default password must be changed before the gateway is connected to the cloud.
Mistake 3: No local historian
Organisations that move the historian exclusively to the cloud lose process visibility the moment connectivity is lost. In OT, operational decisions are made within seconds, so a lack of access to historical data can lead to incorrect operator responses. A hybrid historian with a local copy and cloud replication should be used.
Mistake 4: No monitoring of gateway-cloud traffic
Communication between IIoT gateways and the cloud platform needs monitoring. Without it, an attacker can exfiltrate process data or modify parameters undetected. Network traffic monitoring (NDR) with OT protocol decoding should also cover gateway-cloud traffic.
Mistake 5: IoT devices connected without the security team’s knowledge
The maintenance department adds a sensor with a WiFi module, the energy team installs an LTE modem for generator monitoring, and a vendor connects its diagnostic device. Each of these devices creates a new path into the OT network that the security team does not know about. Active and passive inventorying must cover IoT and IIoT devices.
Incidents from 2024-2025
FrostyGoop (January 2024): The attackers sent Modbus TCP commands to district heating controllers, downgraded their firmware and caused incorrect measurements. Over 600 apartment buildings lost heating, and remediation took almost two days. According to Dragos, the attackers possibly gained access to the network months earlier through an undetermined vulnerability in an externally facing router. The router, management servers and controllers were not adequately segmented. A detailed analysis is in the attack encyclopedia - FrostyGoop.
IOCONTROL (December 2024): Claroty described malware linked to the CyberAv3ngers group, believed to be part of the Iranian IRGC-CEC. It targets IP cameras, routers, PLCs, HMIs and firewalls running Linux, and receives commands over the MQTT protocol.
ICONICS (March 2025): Unit 42 described five vulnerabilities in locally installed SCADA software for Windows, which is used, among other things, for cloud and IIoT integration. Several dozen such servers were reachable from the internet.
Mars Hydro (February 2025): An unprotected database of a manufacturer of home IoT grow lights exposed 2.7 billion records and over 1 TB of data, including Wi-Fi network names and passwords, IP addresses and device IDs.
Cloud-OT risk assessment per IEC 62443
IEC 62443-3-2 defines the process of dividing a system into zones and conduits and assigning them a target security level (SL-T). For cloud-OT deployments this process requires extension to cover cloud specifics.
Step 1: Identify the cloud-adjacent zone
Every IIoT gateway, edge server, or cloud integration point creates a new zone or extends an existing one. The SL-T of this zone follows from the risk assessment, not from the cloud connection alone: what counts is the consequence of a breach in the connected systems (step 2) and the capability of the likely adversary. An IIoT gateway links the process network to the public cloud, so it is usually more exposed than a zone without such a link, and the risk assessment has to account for that.
Step 2: Determine the consequences of compromise
For each cloud-OT connection, the following questions must be answered:
- What happens if an attacker gains access to the cloud platform? (confidentiality)
- What happens if process data in the cloud is modified? (integrity)
- What happens if the cloud service is unavailable for 24h / 72h / 7 days? (availability)
The answers, together with the assessment of adversary capability, determine the target SL-T. The threshold of losses or danger to people that justifies a higher level is set by the asset owner in the risk assessment; the standard does not prescribe it.
Step 3: Describe communication conduits
Every cloud-OT connection is a conduit in the IEC 62443 sense. For each conduit, the following should be recorded: access points, the data flows through each of them (protocol and direction), the SL-T, and requirements and policies, including authentication and logging.
More on the practical implementation of IEC 62443, including zones and conduits, in the dedicated guide.
Regulation and the supply chain
Under NIS2, essential and important entities must manage supply chain risk, including the risk on the side of cloud service providers.
Where we help
SEQRED helps industrial organisations with security audits of cloud-OT architecture, penetration tests covering the cloud-gateway-process network path, and designing secure architectures aligned with IEC 62443.
Sources
- Claroty - State of CPS Security 2025: OT Exposures
- Claroty - One in Four CPS-Enabled Organizations Lost Over USD 1M from Cyber Attacks in Last Year - Claroty (accessed: )
- Claroty - 38% of Riskiest Cyber-Physical Systems Overlooked by Traditional Vulnerability Management - Claroty (accessed: )
- PwC - Global Digital Trust Insights 2026: OT, IIoT and Talent Gaps - Industrial Cyber (accessed: )
- IoT Analytics - State of IoT 2025: Number of connected IoT devices - IoT Analytics (accessed: )
- Dragos - Intel Brief: FrostyGoop ICS Malware - Dragos (accessed: )
- Claroty Team82 - Inside a New OT/IoT Cyber Weapon: IOCONTROL - Claroty (accessed: )
- Palo Alto Unit42 - Multiple Vulnerabilities in ICONICS SCADA - Palo Alto Networks Unit 42 (accessed: )
- ThaiCERT - 1.17TB data leak exposes Wi-Fi passwords and IPs from Mars Hydro IoT grow light devices - ThaiCERT (accessed: )
- Amazon S3 - Default encryption FAQ - AWS (accessed: )
- Directive (EU) 2022/2555 (NIS2) - EUR-Lex (accessed: )
- CSA - Top Threats to Cloud Computing 2025
- Honeywell - Google Cloud and Honeywell: Next Chapter of Industrial Cybersecurity
- techUK - Why You Should Rethink OT Security for the Cloud Era
- Industrial Cyber - Rising ICS Incidents Drive Shift to Intelligence-Driven OT Security
- Waterfall Security - Learning From 2024's Top OT Attacks
- ENISA - Energy Sector Cybersecurity
Need help in this area?
Our experts will help you assess the risk and plan next steps.