Umeå University, the Department of Computing Science, is seeking candidates for a postdoc position in resource-frugal federated learning for preserving security and privacy with focus on edge infrastructures.
The rapid increase of autonomous systems and applications are providing challenges in dealing with petabytes of data. These size and multidimensional features make the machine learning models larger and more complex. Classical centralized approaches to learning and inference fail to address the problems of resource and storage limitations, network bandwidth constraints, tail latency, energy-efficiency, and many more. This project focuses on design and implementation of resource-frugal and robust federated learning algorithms for preserving security and privacy, which are ideally suited for big-data and edge infrastructures.
This project leverages federated learning techniques for advancing state-of-the-art machine learning algorithms where data is geographically distributed and sensitive. Federated learning algorithms empower large-scale distributed nodes, i.e., mobile devices to train globally shared models without divulging the privacy of raw data. Sophisticated attackers leverage the limitations of data, model, target class(es), resources, the communication path for the deception of federated learning algorithms and also to violate security and privacy. By creating unique features (e.g., decentralized optimization, heterogeneity, cost-effective communication architecture, model agnostic learning and robustness) of federated learning algorithms, this project addresses the problems of limited resources, computation, communication, and energy-efficiency for preserving security and privacy. As a result, these features improve the safeguard of services and diagnosis ability of edge infrastructures.