Resource-aware IoT with Enhanced Intelligence and Security

The pervasion of the Internet of Things (IoT), which connects numerous sensors, actuators, appliances, and vehicles, will strongly impact the evolution of more intelligent and greener cities and environmental monitoring. A basic tenet underlying all critical functionalities of the IoT is situational awareness, that is, the ability to capture events and derive accurate, critical information to facilitate decision-making to enable timely action in a heterogeneous and highly dynamic environment. This requires an intelligent infrastructure that is autonomous, dependable, and resilient to natural or artificial disturbances. A critical component of such infrastructure comprises myriads of information-gathering sensors deployed throughout many points of concern in the city. The sensors deployed in the IoT of smart cities constitute critical data sources on which the ensuing analytics and control actions depend. These sensors are interconnected through the internet, forming an essential part of the IoT, and are most likely powered only by batteries. To ensure that the sensors function effectively, we take a holistic approach to designing secure sensor networks with energy-efficient practical algorithms, starting with sensing, followed by data processing and communication to ensure reliable decision-making to enable timely actions that enable long-lasting, secure, and dependable functionality. This project aims to go beyond state-of-the-art solutions and take a holistic approach that starts with smart sensors, intelligent inference, and secure two-way communication among all devices in the network. During the first three years of the project, we developed new distributed machine learning algorithms for estimation and control tasks that reduce the amount of communication in IoT/CPS. We also accounted for system malfunctioning or security threats, for example, noisy and spiky sensor readings or data falsification. In particular, we developed inference methods to promptly identify malicious data tampering and privacy-preserving solutions to avoid information leakage when devices collaboratively solve tasks. In the penultimate year, we implemented efficient scheduling algorithms for channel-constrained remote estimation scenarios utilizing the age-of-information, which indicates the "freshness" of information. We also focused on designing privacy-preserving distributed learning algorithms that protect individual network nodes from leaking private information to internal and external adversaries. Last year, we developed robust methods for distributed algorithms that are resilient to artificial malicious activities by enforcing smoothness across network devices. In 2023, advancements were made in the field of distributed learning and optimization in systems without data transfer to a central hub, involving peer-to-peer interactions and coping with resource constraints such as computational resources, battery power, communication bandwidth, and privacy protection. The findings also underscore the relevance of this research to machine learning problems, such as robust phase retrieval, blind deconvolution, biconvex compressive sensing, and dictionary learning, highlighting its potential impact on various real-world applications.

This project develops efficient detection and estimation schemes to improve data quality and security of the physical-layer signals in IoT. The project contributes to ICT as an enabling technology, and has a strong international component; students hired will spend part of the studies at University of Notre Dame, U.S.A. and at Aalto University, Finland. The pervasion of the Internet of Things (IoT) which connects numerous sensors, actuators, appliances, vehicles etc, will have a strong impact on the evolution of smarter and greener cities as well as on environmental monitoring. A basic tenet underlying all key functionalities of the IoT is situational awareness, i.e., the ability to capture events and derive accurate critical information to facilitate decision making to enable timely action in a heterogeneous and highly dynamic environment. This calls for an intelligent infrastructure that is autonomous, dependable, and resilient to natural or man-made disturbances. A critical component of such an infrastructure comprises myriads of information-gathering sensors deployed throughout many points of concerns in the city. The sensors deployed in smart cities' IoT constitute critical data sources, on which the ensuing analytics and control actions depend. Those sensors are interconnected through the internet, forming an important part of IoT, and most likely powered only by batteries. To ensure that the sensors function effectively, we need to take a holistic approach to designing secure sensor networks with energy-efficient functional algorithms starting with sensing, followed by data processing and communication to ensure reliable decision making to enable timely actions that make possible long lasting secure and dependable functionality. This project aims to go beyond state-of-the-art solutions and take a holistic approach that starts with smart sensors, smart inference, and secure two-way communication among all the devices in the network.