Abstract
Increasingly, smart computing devices, with powerful sensors and internet connectivity, are being embedded into all new forms of infrastructure, from hospitals to roads to factories. These devices are part of the Internet of Things (IoT) and the economic value of their widespread deployment is estimated to be trillions of dollars, with billions of devices deployed. Consider the example of "smart meters" for electricity utilities. Because of clear economic benefits, including a reduction in the cost of reading meters, more precise information about outages and diagnostics, and increased benefits from predicting and balancing electric loads, such meters are already being rolled out across North America. With residential solar collection, smart meters allow individuals to sell power back to the grid providing economic incentives for conservation. Similarly, smart water meters allow water conservation in a drought. Such infrastructure upgrades are infrequent (with smart meters expected to be in service for 20-30 years) but the benefits from the upgrade justify the significant cost. A long-term benefit of such upgrades is that unforeseen savings might be realized in the future when new analytic techniques are applied to the data that is collected. The same benefits accrue to any infrastructure that embeds increased sensing and actuation capabilities via IoT devices, including roads and traffic control, energy and water management in buildings, and public health monitoring.
Introduction
As we transition from "dumb" to "intelligent" infrastructure, the traditional boundary between cybersecurity and physical safety is rapidly blurring. The integration of sensors and actuators into critical systems—such as power grids, water management, and transportation—promises immense economic and efficiency gains. However, this connectivity also introduces a new class of vulnerabilities. This paper outlines a national research agenda to secure the trillions of dollars in IoT-enabled systems that will underpin future cities, emphasizing the need to move beyond traditional software security to address long-term hardware lifecycles and complex system interdependencies.
The Intelligent Infrastructure Gap
The core challenge lies in the "intelligent infrastructure gap." Critical systems are being upgraded with connectivity they were never originally designed to handle. While these upgrades enable features like load balancing in smart grids, they introduce three distinct tiers of threats:
- Individual Device Attacks: These include physical ransomware, such as remotely disabling a home's power, or the theft of sensitive personal data.
- Network-Level Exploits: "Man-in-the-middle" attacks can allow adversaries to control devices by intercepting communications, bypassing the need to compromise the hardware itself.
- Emergent Collection Threats: Large-scale botnets, exemplified by the 2016 Mirai attack, can weaponize thousands of identical devices to cripple the internet backbone or destabilize the power grid.
Compounding these risks is the lifecycle mismatch: infrastructure is built to last 20–30 years, far exceeding the typical 3–5 year support cycle of the technology companies producing the "smart" components.
A Multi-Disciplinary Research Framework
To address these challenges, a shift toward high-assurance engineering is required. The Computing Community Consortium (CCC) proposes a framework centered on several key methodologies:
- Formal Verification: Utilizing mathematical specifications to prove that software is "unattackable." Successes in programs like DARPA’s HACMS, which secured drone software, serve as a proof of concept for this approach.
- Graceful Degradation: Systems must be designed to fail partially rather than suffer total collapse. If a smart component is compromised, the core infrastructure should remain functional in a "safe mode."
- Compartmentalization: Borrowing the "watertight compartment" principle from shipbuilding, digital systems should be segmented to prevent a single failure or hack from cascading across the entire network.
- Adversarial Thinking: Bridging the gap between computer scientists, who focus on data integrity, and traditional engineers, who focus on physical safety, is essential to anticipate how digital exploits translate into physical hazards.
Strategic Insights for Long-Term Security
Several unique problems arise in the context of long-lived infrastructure:
- The "Zombie Device" Problem: When a technology provider goes bankrupt, the sensors they installed in a bridge or road may remain for decades without security patches, becoming permanent, "orphaned" vulnerabilities.
- Physics as a Security Layer: The physical world can serve as a ground truth. If sensor data contradicts the laws of physics, the system can automatically detect a digital compromise through physical reality.
- Privacy via Differential Privacy: To balance data analytics with individual rights, the paper advocates for mathematical guarantees like differential privacy. This allows researchers to extract value from aggregate data without compromising the identity of individuals.
- Human-in-the-Loop: Critical infrastructure requires human oversight. Research is needed to reduce the "complexity burden" on operators, ensuring they can effectively intervene during a crisis without being overwhelmed by data.
Conclusion and Future Directions
The greatest danger to our future infrastructure lies in unforeseen interdependencies—where a failure in a building's lighting system might unexpectedly trigger a shutdown in its HVAC or security systems due to shared network logic. To mitigate these risks, future work must focus on workforce development, training a new generation of operators skilled in both civil engineering and cybersecurity. Additionally, the federal government must find ways to incentivize the private sector to prioritize long-term security over rapid deployment and develop robust methods for pushing security patches to low-resource devices that may remain in the field for thirty years.
By addressing these research priorities, we can ensure that the transition to intelligent infrastructure is as safe as it is transformative.
Summary updated