Secure Federated Edge Intelligence Framework for AI-driven 6G Applications

The state of Qatar has recently announced its national Artificial Intelligent (AI) strategy blueprint identifying AI technologies as a strategical objective in infrastructure development and diverse business landscape. As a revolutionary technology that can serve the entire society, AI is increasingly integrated across a host of applications, including virtual and augmented reality, connected healthcare diagnostics, autonomous driving, and holographic projection. One of the key factors driving this revolution and fueling the trend for adopting AI more broadly is the ever-growing amounts of data and information that can be harvested from a growing number of AI sensors and cameras built into smart devices. Towards a full implementation of AI-based society, a comprehensive digital ecosystem will be pushing/collecting more intelligence to/from a massive number of connected heterogeneous end-devices and systems. However, to support the massive number of users’ connectivity transmitting at the multi-gigabits rate, the network infrastructure should be embedded with a high level of intelligence to efficiently and automatically manage its resources in satisfying diverse Quality of Service (QoS) requirements and meet rising traffic demands. The network should also map the demanded capacity of each AI service to the most suitable Radio Access technologies (RATs). Unfortunately, the advanced specifications of the 5G network, as defined by the 3rd Generation Partnership Project (3GPP) and the International Telecommunication Union (ITU), are not enough to meet these new AI-driven application requirements. Accordingly, both the industry and academia are pushing the research efforts to envision the next generation networks, or what is termed as 6G, to be the enabler technology piloting the evolution of AI-driven applications by upgrading the network from “connected things” to “connected intelligence”. To this end, this research proposal aims to promote the development of new technologies in Qatar by contributing to the success of AI in tackling the challenges of AI-Driven 6G applications. Edge machine learning (ML), as a promising AI component, refers to the development of fast learning models by considering the large distributed computation and data resources available at the network edge. Federated Edge Learning (FEEL), is a nascent edge learning framework for data privacy preservation. FEEL allows training machine learning models by aggregating local learning models at the edge servers instead of users’ raw data to preserve users’ privacy. FEEL requires multiple training iterations where during each iteration round, smart devices locally train their local models using their local data and send these models to the edge server for aggregation to build a global model. Then, the global model is sent back to the end devices, and this whole process is repeated until convergence. With thousands of end devices continuously updating and sending their local models to the edge, network constraints, such as bandwidth underutilization, latency, and spectrum scarcity, can severely affect and degrade the performance of FEEL-based applications. In addition, the unlabeled, heterogeneous, and/or unbalanced data across network participants, when used for training local models, can significantly influence the quality and slow-down the convergence of the global model. These network and application constraints need an efficient network orchestrator that optimizes the utilization of the available network resources while considering the requirements of AI-driven applications. The next-generation 6G networks should be embedded with sufficient intelligence to (i) understand the requirements of FEEL-based applications, (ii) continuously measure key network performance metrics, and (iii) automatically learn and decide the optimal resource allocation strategy, therefore ensuring a high QoS for FEEL-based applications. Furthermore, dealing with the security concerns that can harm both users’ privacy and learning performance is of paramount importance for the successful deployment of the future AI-Driven 6G applications. In this project, a Secure Federated Edge Intelligence Framework will be developed to respond to the requirements of FEEL in regards of network, data, and spectrum constraints, as well as the security and privacy concerns. This framework will fuel the AI industry in Qatar to thrive with the emergence of 6G networks. The proposed framework aims to achieve the objective of mitigating the challenges mentioned above through the following three sub-objectives, Obj1) developing reliable and resource-efficient edge intelligence algorithms to optimize FEEL experience under the network, and data constraints, Obj2) efficiently managing radio frequency (RF) spectrum resources by considering limited knowledge of the environment and enabling spectrum trading for resource exchange and better spectrum utilization, and Obj3) securing the FEEL and enhancing the privacy of the users. As a proof-of-concept, developed FEEL algorithms will be tested against real-world datasets, and a software-defined radio-based prototype will be implemented for smart spectrum management based on spectrum measurements obtained by the aid of Ooredoo and TASMU stakeholders. A private blockchain network will be deployed to serve as the trading platform, as well as providing security and privacy enhancement for FEEL.

Hamad Bin Khalifa University (HBKU)