In recent years, the use of Artificial Intelligence (AI) has become widespread, leading to a technological paradigm shift. Addressing some of the main objectives outlined in the United Nations' Sustainable Development Goals (SDGs) for 2030 necessitates the responsible use of AI techniques to convert data into actionable knowledge for societal benefit. This trend is propelled by increasing hyperconnectivity through the integration of distributed systems into Internet infrastructure, particularly leveraging Internet of Things (IoT) technologies and 5G/6G infrastructures. The integration of such systems enables new data-based services, such as those in sustainable cities, communities, and advanced eHealth services. REMINDER aims to design a decentralized and secure approach for accessing and processing data from distributed systems, including an edge-based architecture for Federated Learning (FL). This architecture facilitates collaborative model creation without sharing raw data, addressing major security and privacy challenges associated with decentralized Machine Learning (ML) approaches like FL. The project also focuses on data privacy compliance with regulations like GDPR, addressing security attacks, and demonstrating feasibility through eHealth and smart building use cases.
REMINDER focuses on addressing key security and privacy challenges in decentralized and distributed systems. Through the integration of advanced technologies such as Federated Learning (FL) and cryptographic approaches, REMINDER aims to develop secure and privacy-respecting methods for managing the data lifecycle.
REMINDER is designing new authentication protocols to ensure exclusive participation of legitimate systems during decentralized training processes.
The project will explore advanced use cases aligned with the United Nations' Sustainable Development Goals (SDGs), focusing on the application of FL in eHealth and energy efficiency.
REMINDER will propose an edge-enabled architecture where edge nodes will act as auxiliary components to enable the integration of constrained devices into the architecture.
The project will design a client selection approach to identify the most suitable nodes to participate in the FL training process, considering the security level provided by different devices.
REMINDER will design a privacy-preserving framework for using FL in distributed systems.
REMINDER aims to contribute to building a safer and more privacy-respecting digital future by addressing critical challenges in data management in decentralized and distributed environments.
VISION AND AMBITION IN SCIENCE-ENABLED TECHNOLOGY
Addressing the opportunities and risks of AI-driven, data-centric societies through interconnected IoT and 5G/6G networks, REMINDER focuses on enhancing cybersecurity and privacy within decentralized systems.
COMPREHENSIVE RESEARCH APPROACH
REMINDER seeks to overcome privacy and performance concerns in ML deployments by exploring Federated Learning (FL), integrating blockchain, employing Differential Privacy (DP), and addressing dynamic system behaviors across eHealth and smart buildings applications.
INTERDISCIPLINARY COLLABORATION
By leveraging expertise in cybersecurity, edge computing, and ML, REMINDER's consortium, including UMU, UWE, AIT, and SIE, demonstrates interdisciplinary collaboration to solve complex challenges in decentralized systems.
EVALUATING RISKS AND FLEXIBILITY
REMINDER addresses computational demands on IoT devices through model pruning and compression, explores lightweight ML models, and remains flexible in adapting to technical findings and the needs of planned use cases while mitigating risks in privacy-preserving FL techniques.
REMINDER’s consortium comprises four esteemed partners from diverse sectors across four countries: Spain, UK, Romania, and Austria. Each partner brings unique expertise and experience, collectively forming a purposeful and multidisciplinary team. With a solid track record in research and innovation, these partners embody a commitment to advancing security and privacy in decentralized systems.
