Cybersecurity
Threat Modeling Connected Sensors: A Practitioner's Guide to Securing the Edge
A step-by-step methodology for embedded engineers and OT practitioners to identify, assess, and mitigate security risks in constrained sensor deployments without over-engineering solutions.

Why sensor security matters in modern OT environments
Operational technology (OT) environments have traditionally been air-gapped or isolated, but the drive toward digital transformation, predictive maintenance, and remote management has introduced connected sensors at the edge. These devices—ranging from vibration monitors on rotating equipment to environmental sensors in clean rooms—often communicate over industrial protocols such as Modbus, PROFINET, or MQTT. While they generate valuable data, they also expand the attack surface. A compromised sensor can serve as a foothold for lateral movement, feed false data to control systems, or become a persistent presence in the network. For embedded engineers and OT practitioners, understanding how to model threats specific to these devices is the first step toward protecting critical infrastructure.
Unlike traditional IT assets, sensors are frequently constrained by limited processing power, memory, and energy. They may run firmware that is years old, lack built-in update mechanisms, or rely on default credentials. These characteristics make conventional security controls—such as heavy endpoint agents or real-time intrusion detection—impractical. Threat modeling for this context must be lightweight, focused on the unique trust boundaries of the edge, and aligned with the operational realities of the shop floor or field deployment.
Asset inventory and trust boundary definition
Before modeling threats, practitioners must establish what is being protected. In a sensor network, the asset inventory includes not only the physical device but also the communication modules, configuration stores, and any attached actuators. Each sensor typically reports to a gateway or edge node, which in turn relays data to a central historian or cloud platform. The trust boundary is defined by where control transitions from an unmanaged network—such as a wireless fieldbus—to a managed segment, such as a demilitarized zone (DMZ) or plant DMZ.
A practical approach begins with a simple data-flow diagram. Identify every sensor, its communication protocol, and the endpoint it contacts. Note the direction of data flow: does the sensor push data periodically, pull commands on demand, or operate in a hybrid mode? Annotate which components are under your control and which are operated by third-party vendors or legacy systems. This inventory becomes the foundation for the threat model; without it, any assessment is speculative. Importantly, document any hard-coded credentials, unencrypted communication channels, or outdated firmware versions that exist by design rather than oversight.
Common threat vectors for connected sensors
Connected sensors face a distinct set of threat vectors that differ from general IT threats. Eavesdropping is prevalent when sensors transmit over unencrypted radio links or legacy serial protocols. An attacker within radio range can capture sensor readings and, in some cases, inject false commands. Replay attacks are also common: if a sensor accepts authenticated messages without nonce validation, an attacker can capture a legitimate command—such as a set-point change—and replay it to cause unintended behavior.
Firmware tampering represents another significant risk. Sensors that accept over-the-air (OTA) updates without integrity verification can be persuaded to run malicious code. Once executed, this code can disable reporting, exfiltrate data, or pivot to other devices on the same network. Default or hard-coded credentials are a persistent issue; many sensors ship with passwords like "admin/admin" that are never changed, providing immediate access to anyone who consults the device manual. Finally, physical access to a sensor—whether in a cabinet or mounted on a machine—enables adversaries to extract firmware, reset the device to factory defaults, or attach debugging interfaces.
Practical threat modeling techniques
Threat modeling need not be an exhaustive process. For sensor deployments, a streamlined methodology based on the STRIDE framework adapted for OT can be effective. Start by listing each component identified in the asset inventory and ask STRIDE-related questions: Does this component allow Spoofing of identity? Can data be Tampered with in transit? Does the design lead to Information disclosure? Could an attacker Repudiate actions? Does the sensor Deny service to legitimate users? Is there an Elevation of privilege possible?
A useful practical exercise is the creation of an attack tree for a representative sensor. Begin with a high-level goal—such as "manipulate temperature reading"—and break it into substeps. Each step reveals a required capability, such as "intercept radio traffic" or "extract firmware binary." This structure helps prioritize mitigations that address the most likely or highest-impact paths. For teams new to threat modeling, Microsoft's Threat Modeling Tool or even a structured spreadsheet can serve as the repository for findings, provided the output is kept lightweight and focused on actionable items.
Mitigation strategies within hardware constraints
Mitigations for sensor security must respect the reality that many devices cannot run heavyweight security agents or support full TLS stacks. Where encryption is required, consider lightweight alternatives such as DTLS with pre-shared keys or AES-CCM modes that operate within limited RAM. If the sensor supports OTA updates, mandate signed images and hash verification before installation; a simple checksum can prevent the execution of corrupted or malicious firmware.
Network-level controls are often more feasible than host-level ones. Placing sensors behind a gateway that performs protocol validation can strip malformed packets before they reach the device. Firewall rules that restrict traffic to known ports and IP addresses reduce the attack surface. For wireless sensors, frequency hopping or spread-spectrum techniques can make eavesdropping more difficult, though they provide obscurity rather than strong confidentiality. When default credentials exist, they must be rotated during commissioning, and a policy should be enforced that prohibits hard-coded passwords in future deployments.
Physical security is equally important. Where possible, sensors should be mounted in enclosures that prevent unauthorized access, and any debug interfaces (such as UART or SWD) should be disabled or protected by a physical jumper. For deployments where physical access is unavoidable, consider tamper-evident seals or sensors that zeroize keys upon detection of casing opening.
Verification and continuous improvement
A threat model is only as good as its ability to guide actual security improvements. Verification should begin with a review of the implemented controls against the documented threats. Check that encryption is active where required, that firmware signing is enforced, and that network segmentation is in place. Packet capture tools can be used to confirm that communication flows as expected and that no unencrypted sensitive data leaves the sensor.
For ongoing improvement, integrate the threat model into the device lifecycle management process. When a firmware update is released, re-run the threat model to assess whether new vectors have been introduced. For fleets of sensors, consider automated inventory tools that can detect configuration drift—such as a sensor reverting to default settings after a power cycle—and trigger a re-assessment. Documentation should be versioned; as the deployment evolves, the threat model must evolve in step. Remember that the goal is not to eliminate all risk—that is often impossible in OT environments—but to reduce risk to an acceptable level and maintain that posture over time.
Practical checklist for sensor threat modeling
- List all sensors and their communication paths; document protocol, endpoints, and data types.
- Define trust boundaries between field devices, gateways, and central systems.
- Identify and catalog common threat vectors: eavesdropping, replay, firmware tampering, default credentials, physical access.
- Apply a simplified STRIDE analysis to each component, focusing on spoofing, tampering, and information disclosure.
- Verify that OTA updates require signed images and integrity checks.
- Enforce network segmentation and firewall rules limiting sensor traffic to necessary endpoints.
- Confirm physical hardening: enclosures, disabled debug interfaces, tamper detection where feasible.
- Schedule periodic reviews of the threat model when firmware, topology, or vendor contracts change.
Explore Rootcastle engineering tools → · Public profiles and publications


