Connected embedded devices are becoming integral to industrial operations and other critical services. As these systems exchange data and interact with physical processes, a security incident can affect more than information confidentiality: it may disrupt operational capability. Cyber resilience in embedded systems therefore requires more than adding protective features. It calls for an approach that identifies risks, detects suspicious activity, supports an appropriate response, and helps restore the system while preserving useful evidence.
For automation engineers, system integrators, and plant leaders, this shifts security from a collection of isolated controls to a lifecycle concern. Device manufacturers and their partners need to consider how security is designed, managed, and maintained across development and deployment—and what the system can do if its protections are compromised.
Cyber resilience in embedded systems: beyond prevention
Cyber resilience is the ability of a system to identify, prevent, and respond to cyberattacks that could disrupt its operation while maintaining data confidentiality and integrity. In practice, this means preparing for both prevention and failure: a defence may be bypassed, a software vulnerability may emerge, or a component may behave in an unexpected way.
Security frameworks provide useful functions for structuring this work. The NIST Cybersecurity Framework identifies five core functions: identify, protect, detect, respond, and recover. Related guidance emphasizes managing risk, protecting systems, discovering incidents, minimizing their impact, and learning from them. These functions help connect organizational security objectives to embedded-device requirements.
They also highlight why prevention alone is insufficient. Many embedded security approaches rely heavily on establishing and maintaining trust—for example, through secure boot, cryptographic protections, or isolated execution environments. These mechanisms can provide important protections, but a design that depends on trust needs to consider what happens when that trust is broken. A reboot or reset may restore operation, yet it may not reveal what occurred or provide the information needed to improve the response.
Translate security functions into device requirements
Regulations and frameworks provide direction, but they do not automatically specify how an embedded device should behave. Manufacturers and system designers must translate high-level security objectives into requirements suited to the device, its use case, and its deployment environment.
Identify and protect
Identification begins with understanding the application and deployment scenario. Designers can decompose the system, examine interactions among components and external entities, and assess potential threats and their consequences. Threat and security modelling helps make assumptions about an attacker, possible goals, and potential methods explicit. The resulting assessment can guide which risks deserve priority and which controls are appropriate.
Protection then applies safeguards for information and system resources. Relevant objectives include confidentiality, integrity, availability, and authentication. Cryptographic methods can help protect data and verify identity or integrity, while isolation mechanisms can separate resources or execution environments. Such controls are foundational, but they do not replace ongoing monitoring or an incident plan.
Detect, respond, and recover
Detection involves monitoring relevant system activity and looking for behaviour that differs from expected or healthy operation. When a potential incident is identified, the system needs a way to report it and support analysis. Response planning determines how to limit the impact, while recovery may involve repair, updates, or returning the device to a known healthy state.
Recovery is stronger when it is informed by evidence. Without system-level visibility, it can be difficult to understand the cause and method of a failure, assess its scope, or refine future mitigations. A useful resilience strategy therefore considers both restoration and the information needed to evaluate the incident.
Why trust-based security needs operational support
Embedded security mechanisms have often been designed to address known attack classes or provide passive safeguards. Examples include monitoring voltage or temperature, using watchdog timers, or resetting a system. These measures can have a role, but they may not provide an active response tailored to a detected threat. Security updates can address some software issues, but weaknesses built into a device’s microarchitecture may be more difficult to correct after release.
Trust-based designs also depend on the integrity of multiple components and assumptions. If one link is compromised, protections that rely on it may no longer be dependable. Research discussed in the source paper identifies examples of weaknesses affecting trusted software, processor isolation, and system communication. The practical lesson is not that any single control is ineffective, but that security architecture should account for the possibility that a control may fail.
For industrial deployments, this makes clear ownership important. Device security should not be treated solely as a feature delegated to a third party. Risk assessment, secure design practices, and security maintenance need defined responsibilities across the device lifecycle and supply chain.
Three architectural characteristics for resilience
A proposed resilient embedded architecture can build on existing protections with three complementary characteristics:
- An independent active runtime system security manager: This component would monitor system resources, assess gathered information, and coordinate protection, detection, response, and recovery. The proposal calls for physical independence and isolation from the general-purpose processor and its memory resources, reducing reliance on the same resources being monitored.
- Active runtime resource monitors: Resource-specific monitors would report fine-grained information about activity to the security manager. That visibility can help distinguish normal from suspicious behaviour and provide information for incident analysis.
- An active response manager: This component would enforce response and recovery strategies initiated by the security manager. Depending on the architecture, a compromised resource could be isolated, allowing the system to limit an incident and potentially preserve critical functions.
These are proposed architectural characteristics, not a claim that every current embedded platform provides them. Their purpose is to make security activity observable and to support action when preventive controls are no longer enough. Continuous monitoring and event records can also help maintain a history of system activity for evaluating a breach and informing future mitigation.
Practical implications for industrial teams
When evaluating embedded devices or planning a system design, technical teams can use the five security functions as a practical review structure. Ask what assets and interactions have been assessed, how information and resources are protected, what activity is monitored, how incidents are handled, and how the system returns to a healthy state. Also consider what evidence will be available to understand an incident.
These questions are relevant across product development and deployment. Integrators and operators can clarify security responsibilities with manufacturers and service partners, while manufacturers can incorporate security requirements into design and lifecycle processes. The specific controls will depend on the system and its risks; the goal is to connect them into a coherent plan rather than rely on a single defence.
Conclusion
Cyber resilience in embedded systems means planning for compromise as well as prevention. For critical infrastructure, the practical objective is to combine risk management and protection with runtime visibility, an actionable response, and recovery informed by evidence. An architecture that supports independent security management, resource-level monitoring, and active enforcement offers a way to address gaps in passive, trust-dependent approaches. Treating these capabilities as lifecycle requirements can help industrial teams make security decisions that account for both device behaviour and operational continuity.