A Survey of Robustness in Trustworthy Federated Learning Against Malicious Attacks

PhD Qualifying Examination


Title: "A Survey of Robustness in Trustworthy Federated Learning Against 
Malicious Attacks"

by

Mr. Xingxing TANG


Abstract:

The advent of stringent data protection regulations such as the General Data 
Protection Regulation (GDPR) has imposed significant challenges in the realm of 
data utilization and machine learning. Federated Learning (FL) has accordingly 
gained traction as a privacy-preserving machine learning paradigm that 
circumvents the direct sharing of private data across various users or 
organizations. Despite its promise, FL is not immune to vulnerabilities; it 
encounters malicious attacks that could degrade the performance and implant 
Backdoors into the federated models. This survey examines the landscape of 
Trustworthy Federated Learning (TFL) with a focus on the robustness of such 
systems against malicious attacks.

In this survey, we establish a unified framework for the Definition of 
Robustness Guarantee Against Malicious Attacks in TFL. We provide a taxonomy of 
known malicious threats and dissect contemporary defense mechanisms designed to 
fortify FL systems. Furthermore, we delve into the intricate balance between 
robustness, privacy, and utility, achieved through multi-objective optimization 
in FL.

This survey encapsulates the nascent yet critical discourse on ensuring the 
Robustness of FL systems, offering scholars and practitioners a lens through 
which to assess and enhance the robustness of FL. We conclude with future 
directions, underscoring the need for innovative approaches that balance 
robustness, privacy, and utility in FL.


Date:                   Monday, 15 April 2024

Time:                   10:00am - 12:00noon

Venue:                  Room 4475
                        Lifts 25/26

Committee Members:      Prof. Qiang Yang (Supervisor)
                        Prof. Kai Chen (Co-supervisor)
                        Prof. Bo Li (Chairperson)
                        Dr. Dongdong She