Title: Advancements in Deep Learning for Graph-level Anomaly Detection

Abstract:

The capacity to capture intricate relationships among diverse entities positions graphs as indispensable and extensively employed representations in real-world applications. Consequently, anomaly detection within graphs finds diverse applications, such as identifying anomalous (e.g., congested) network devices in mobile communication services. Despite the widespread use of graph data and the significance of anomaly detection, there has been a relative lack of focus on Graph-Level Anomaly Detection compared to the detection of anomalous nodes within graphs. In this keynote speech, I will introduce two of our recent works focusing on graph-level anomaly detection. Our first work addresses the challenge of detecting both locally and globally anomalous graphs. The approach leverages deep learning techniques, specifically employing random distillation of graph and node representations. Through the joint training of graph neural networks with randomly initialized weights, this model outperforms seven state-of-the-art methods, showcasing its efficacy in diverse domains. Our second work introduces Hierarchical Memory Networks (HimNet), a pioneering approach employing a graph autoencoder network architecture. HimNet incorporates node and graph memory modules to capture both fine-grained, local anomalies and holistic, global abnormalities. This dual-module system offers robust detection capabilities against anomalies, as demonstrated through extensive empirical results on 16 real-world graph datasets.

Short Bio:

Jin Li is currently a professor at Guangzhou University. He got his Ph.D degree in information security from Sun Yat-sen University at 2007. His research interests include design of secure protocols in Artificial Intelligence, Cloud Computing (secure cloud storage and outsourcing computation) and cryptographic protocols. He has published more than 100 papers in international conferences and journals, including IEEE INFOCOM, IEEE TIFS, IEEE TPDS, IEEE TOC and ESORICS etc. His work has been cited more than 18000 times at Google Scholar and the H-Index is 40. He is Editor-in-Chief of International Journal of Intelligent Systems. He also serves as Associate editor for several international journals, including IEEE Transactions on Dependable and Secure Computing, Information Sciences.

Title: Advancements in Deep Learning for Graph-level Anomaly Detection

Abstract:

The capacity to capture intricate relationships among diverse entities positions graphs as indispensable and extensively employed representations in real-world applications. Consequently, anomaly detection within graphs finds diverse applications, such as identifying anomalous (e.g., congested) network devices in mobile communication services. Despite the widespread use of graph data and the significance of anomaly detection, there has been a relative lack of focus on Graph-Level Anomaly Detection compared to the detection of anomalous nodes within graphs. In this keynote speech, I will introduce two of our recent works focusing on graph-level anomaly detection. Our first work addresses the challenge of detecting both locally and globally anomalous graphs. The approach leverages deep learning techniques, specifically employing random distillation of graph and node representations. Through the joint training of graph neural networks with randomly initialized weights, this model outperforms seven state-of-the-art methods, showcasing its efficacy in diverse domains. Our second work introduces Hierarchical Memory Networks (HimNet), a pioneering approach employing a graph autoencoder network architecture. HimNet incorporates node and graph memory modules to capture both fine-grained, local anomalies and holistic, global abnormalities. This dual-module system offers robust detection capabilities against anomalies, as demonstrated through extensive empirical results on 16 real-world graph datasets.

Short Bio:

Ling Chen is a Professor in the School of Computer Science at the University of Technology Sydney. She leads the Data Science and Knowledge Discovery Laboratory (The DSKD Lab) within the Australian Artificial Intelligence Institute (AAII) at UTS. Ling has been persistently working in the area of machine learning and data mining for 20 years, dedicated to undertaking innovative research to produce high quality results and attracting and leading research and industry projects to initiate and investigate new research areas and create real impacts. Ling's recent research interests include anomaly detection, data representation learning, and dialogues and interactive systems. Her research has gained recognition from both government agencies, receiving competitive grants such as ARC DP/LP/LIEF, and industry partners, with contracted research support from entities like Facebook Research and TPG Telecom. Ling serves as an Editorial Board member for journals including the IEEE Journal of Social Computing, the Elsevier Journal of Data and Knowledge Engineering and the Computer Standards and Interfaces.