Towards a Resilient Federated Edge Intelligence: A Testbed for Design, Analysis, and Validation of Federated Learning
Disciplines
Artificial Intelligence and Robotics | Computer Sciences
Abstract (300 words maximum)
Federated learning (FL) is an efficient, privacy-preserving distributed learning paradigm that enables massive edge devices to train machine learning models collaboratively. Although various communication schemes have been proposed to expedite the FL process in resource-limited wireless networks, the unreliable nature of wireless channels was less explored. In addition, current research on the evaluation of FL is mainly based on the simulation of multi-clients/processes on a single machine/device. However, there needs to be more understanding of the performance of FL under unreliable communication in real-world distributed low-power IoT devices. This research aims to develop a testbed for evaluating FL under unreliable communication. The core of the proposed testbed will constitute Heterogeneous physical devices (e.g., IoT devices) that can be configured to mimic the operation of real FL operations with application software that can be set up to test communications between the devices. The testbed will allow performing effects of different network conditions, such as latency, jitter, packet loss, and bandwidth. The testbed being developed by this project will provide researchers and practitioners with an open and adaptive environment for measurement and experimentation in the FL context. It will also enable opportunities to design and test effective techniques that provide robust FL solutions. In addition, this project can help analyze and validate issues related to FL security.
Academic department under which the project should be listed
CCSE - Computer Science
Primary Investigator (PI) Name
Liang Zhao
Towards a Resilient Federated Edge Intelligence: A Testbed for Design, Analysis, and Validation of Federated Learning
Federated learning (FL) is an efficient, privacy-preserving distributed learning paradigm that enables massive edge devices to train machine learning models collaboratively. Although various communication schemes have been proposed to expedite the FL process in resource-limited wireless networks, the unreliable nature of wireless channels was less explored. In addition, current research on the evaluation of FL is mainly based on the simulation of multi-clients/processes on a single machine/device. However, there needs to be more understanding of the performance of FL under unreliable communication in real-world distributed low-power IoT devices. This research aims to develop a testbed for evaluating FL under unreliable communication. The core of the proposed testbed will constitute Heterogeneous physical devices (e.g., IoT devices) that can be configured to mimic the operation of real FL operations with application software that can be set up to test communications between the devices. The testbed will allow performing effects of different network conditions, such as latency, jitter, packet loss, and bandwidth. The testbed being developed by this project will provide researchers and practitioners with an open and adaptive environment for measurement and experimentation in the FL context. It will also enable opportunities to design and test effective techniques that provide robust FL solutions. In addition, this project can help analyze and validate issues related to FL security.