A Survey on Privacy Problem in Graph Learning

PhD Qualifying Examination


Title: "A Survey on Privacy Problem in Graph Learning"

by

Mr. Qi HU


Abstract:

The proliferation of graph-structured data and the development of graph 
learning techniques, such as graph neural networks, have revolutionized various 
applications. On the one hand, these advancements have enabled the extraction 
of valuable insights from complex graph data, leading to unprecedented 
opportunities for knowledge discovery and decision-making. On the other hand, 
the increasing availability of graph data and the growing capabilities of graph 
learning models have also introduced significant privacy concerns. Despite 
ongoing research efforts to address these privacy risks, the problem remains 
largely unresolved. In this survey, we provide a comprehensive review of 
privacy in graph learning, categorizing privacy attacks according to the 
adversary's motivations to highlight the vulnerabilities present in graph 
data and models. We then present a detailed overview of prominent defense 
strategies that have been developed to mitigate these privacy risks, including 
data anonymization, differential privacy, and private graph learning. Beyond 
existing works, we identify emerging privacy concerns as graph learning 
continues to evolve. Lastly, we outline several promising directions for future 
research to ensure the responsible and privacy-preserving use of graph- 
structured data.


Date:                   Thursday, 16 May 2024

Time:                   10:00am - 12:00noon

Venue:                  Room 3494
                        Lifts 25/26

Committee Members:      Dr. Yangqiu Song (Supervisor)
                        Dr. Wei Wang (Chairperson)
                        Dr. Dongdong She
                        Dr. Binhang Yuan