Title: Resilient Multiple Graph Learning

Abstract: Rapid development of crowdsourced websites and information technology enables us to collect massive amounts of graph data, which are also known as networked data, ranging from biological, bibliographical, knowledge, and social networks, to communication, electrical, geographic, and transportation networks. Multiple graph data analysis has become a powerful tool for gaining insights and deriving innovations into our increasingly connected world. Real-world graph data are typically noisy due to massive disinformation injected by malicious parties and users. Unfortunately, graph learning models, especially deep learning models, are highly sensitive to small perturbations of their input intended to result in analysis failures. Given the need to understand the vulnerability and resilience of graph data analysis, two questions arise: (1) Are multiple graph learning models sensitive to adversarial perturbations over intra-graph and inter-graph interactions? (2) Can we propose impelling defense techniques to offer sufficient protection to multiple graph learning models against adversarial attacks?

In this talk, I will introduce problems, challenges, and solutions for characterizing and understanding and learning vulnerability and resilience of multiple graph learning under adversarial attacks. I will also discuss our recent work on adversarial attacks over multiple graph learning. I will conclude the talk by sketching interesting future directions for resilient multiple graph learning.