What cyber-physical security means
Cyber-physical security is the practice of protecting systems in which computers and networks control physical processes, so that a cyberattack or failure can’t cause physical damage, unsafe conditions or loss of essential services. It covers things like power grids, water treatment plants, factory production lines, vehicles, medical devices and building control systems.
The defining difference from ordinary IT security is the consequence. A breach of an office network might leak data or stop email. A breach of a cyber-physical system can open a valve, overspeed a motor, turn off a safety system or cut power to a city. Security failures become physical events.
Cyber-physical systems work in a loop: sensors measure the physical world, software decides what to do, and actuators act on it. Cyber-physical security has to protect every part of that loop: the sensors’ readings, the controllers’ logic, the network between them, and the commands sent to actuators. For background on the systems themselves, see Cyber-Physical Systems vs Embedded Systems.
Two senses of the term
You’ll see “cyber-physical security” used in two related ways:
- Security of cyber-physical systems (sometimes shortened to CPS security). This is the main meaning and the focus of this article: protecting operational technology (OT), industrial control systems (ICS), connected devices and similar systems.
- Converged physical and cyber security. In corporate security, the phrase can also mean bringing physical security (guards, cameras, badge access, locks) and cybersecurity teams together. Modern cameras and access control systems are networked devices themselves, so an attacker can combine physical and digital routes.
The two overlap. A networked door controller is both a physical security device and a small cyber-physical system.
Why it’s different from IT security
| Aspect | Traditional IT security | Cyber-physical / OT security |
|---|---|---|
| Top priority | Confidentiality of data, then integrity and availability | Safety and availability first, then integrity, then confidentiality |
| Worst outcome | Data breach, financial loss, downtime | Injury, environmental damage, equipment destruction, service outages |
| System lifetime | A few years | Often decades |
| Patching | Frequent, often automated | Difficult: may need a plant shutdown, recertification or vendor approval |
| Timing | Delays are usually tolerable | Control loops can depend on millisecond-level timing |
| Typical devices | Servers, laptops, phones | PLCs, RTUs, sensors, drives, embedded controllers, safety systems |
| Protocols | Mostly modern, with authentication | Many legacy industrial protocols designed without authentication |
Some of the hardest problems come from legacy equipment. Many industrial devices were designed decades ago on the assumption that they’d sit on isolated (“air-gapped”) networks. As companies connect plants to corporate networks and the cloud for monitoring and analytics, those devices face threats they were never built to handle. Some common industrial protocols will accept a command from anyone who can reach the device on the network.
What kinds of systems need it
- Energy: generation, transmission, substations, distributed energy resources
- Water and wastewater: treatment plants, pumping stations, dosing systems
- Manufacturing and process industries: production lines, chemical plants, oil and gas
- Transportation: vehicles, rail signalling, aviation, ports
- Healthcare: infusion pumps, patient monitors, imaging equipment, hospital building systems
- Buildings and cities: HVAC, elevators, access control, lighting, traffic management
- Consumer IoT that controls physical things: smart locks, thermostats, garage door openers
How attacks on cyber-physical systems work
Attackers usually don’t start at the controller. Common paths include:
- Through the IT network. Phishing or a compromised corporate system, followed by movement into the operational network across weak segmentation.
- Remote access. Poorly secured remote-maintenance connections for vendors and technicians.
- Supply chain. Compromised software, firmware or engineering tools.
- Removable media and engineering laptops carried into isolated sites.
- Exposed devices. Controllers or connected devices reachable from the internet, sometimes with default credentials.
- Wireless links that lack proper authentication or encryption.
Once inside, attacks specific to cyber-physical systems include:
- Manipulating sensor data so the controller sees false readings (false data injection or sensor spoofing)
- Sending unauthorised commands to actuators
- Modifying controller logic or firmware so the process behaves unsafely
- Disabling or bypassing safety systems
- Hiding the attack by replaying normal-looking data to operators while the process misbehaves
- Denial of service that breaks the timing of control loops
Real-world incidents
A few well-documented cases show what’s at stake:
- Stuxnet (discovered in 2010) was malware that targeted industrial controllers at Iran’s Natanz uranium enrichment facility. It altered centrifuge speeds while showing operators normal readings. It’s widely cited as the first known cyberweapon designed to cause physical damage.
- The Ukrainian power grid attack (December 2015) cut electricity to roughly 225,000 customers for several hours. Attackers used compromised remote-access tools to open breakers at distribution substations and disrupted recovery efforts.
- TRITON/TRISIS (2017) targeted the safety instrumented system at a petrochemical facility in the Middle East. Safety systems exist to shut a process down safely when something goes wrong, so attacking them is especially dangerous.
- Remote vehicle hacking research (2015). Security researchers showed they could remotely affect a Jeep Cherokee’s systems through its connected infotainment unit. The work led to a recall of about 1.4 million vehicles to fix the vulnerability.
These cases share a pattern. The attacks crossed from digital networks into physical control, and they exploited assumptions that the control network was trustworthy.
Core protections
At the organisation and network level
- Asset inventory. You can’t protect devices you don’t know about. Keep a complete list of controllers, sensors, gateways, firmware versions and connections.
- Network segmentation. Separate the operational network from corporate IT, and divide it into zones with tightly controlled conduits between them. The goal is that a compromised office laptop can’t talk to a controller.
- Strict remote access. Use multi-factor authentication, time-limited access for vendors, logging, and no always-on remote connections that nobody monitors.
- Monitoring designed for OT. Watch for unusual commands, new devices, firmware changes and abnormal process behaviour, not just malware signatures.
- Incident response and recovery plans that include manual operation and safe shutdown.
- Careful patch management, accepting that some systems can only be updated during planned outages.
At the device and firmware level
For engineers designing embedded controllers and connected devices:
- Secure boot so a device runs only firmware signed by its manufacturer
- Signed, rollback-protected updates, as described in Over-the-Air (OTA) Firmware Updates for IoT Devices
- Unique per-device credentials and protected key storage, never shared default passwords
- Authenticated, encrypted communication between devices, gateways and servers
- Locked debug interfaces (JTAG/SWD) in production units
- Input validation on commands and messages, including range checks on setpoints
- Plausibility checks on sensor data, such as comparing redundant sensors or rejecting physically impossible changes
- Minimal attack surface: disable unused services, ports and radio features
Safety and security together
A key principle in cyber-physical design is that safety shouldn’t depend on software being uncompromised. Independent safety layers, such as mechanical relief valves, hardwired emergency stops, independent safety controllers and physical limit switches, can stop a dangerous situation even if the main control system has been taken over. Safety and security engineering increasingly work together, because a security failure can create a safety hazard.
Standards and frameworks
- IEC 62443 is the main international series of standards for industrial automation and control system security. It covers asset owners, system integrators and product suppliers, and introduces concepts like security zones, conduits and security levels.
- NIST SP 800-82 is the US National Institute of Standards and Technology’s guide to operational technology security.
- The NIST Cybersecurity Framework is a general framework that many OT programmes map onto.
- Sector-specific rules apply in many places, such as NERC CIP for the North American bulk power system, and medical device cybersecurity expectations from regulators like the US FDA.
- Product regulation is growing too. The EU’s Cyber Resilience Act sets security requirements for many products with digital elements sold in the EU. We touch on it in New Trends in Embedded Systems and Software Engineering.
FAQ
What’s the difference between OT security and cyber-physical security? They overlap heavily. OT (operational technology) security usually refers to industrial environments: plants, utilities, factories. Cyber-physical security is a broader term that also covers vehicles, medical devices, building systems and consumer IoT that controls physical things.
Is IoT security the same as cyber-physical security? Not exactly. IoT security covers all connected devices, including ones that only collect data. It becomes cyber-physical security when the device can affect the physical world, like a smart lock, an insulin pump or a connected industrial valve.
Why can’t industrial systems just be patched like computers? Many run continuous processes that can’t stop without significant cost or risk. Updates may need vendor testing or safety recertification, and some older devices no longer receive updates at all. That’s why segmentation, monitoring and compensating controls matter so much in OT environments.
Where should an embedded engineer start learning about this? Start with secure boot, firmware signing, key storage and secure communication on a microcontroller platform you already know. Then read an overview of IEC 62443’s component requirements to see how device features fit into a full system’s security.