GTA³ 4.0: The 4th workshop on Graph Techniques for Adversarial Activity Analytics

December 10th, 2020 | Virtual Conference - Atlanta, GA

In addition, we have been informed by the main conference that the event will take place online due to the COVID-19 pandemic.

Theme & Purpose

Networks are natural analytic tools in modeling adversarial activities (e.g., smuggling, illegal arms dealing, illicit drug production) in different contexts. However, such activities are often covert and embedded across multiple domains. They are generally not detectable and recognizable from the perspective of an isolated network, and only become apparent when multiple networks are analyzed in a joint manner. Thus, one of the main research topics in modeling adversarial activities is to develop effective techniques to align and fuse information from different networks into a unified representation for global analysis. Based on the combined network representation, an equally important research topic is on detecting and matching indicating patterns to recognize the underlining adversarial activities in the integrated network. Furthermore, more sophisticated activities could potentially involve attempts at covering their tracks by attacking and changing networks. Thus, a developing area of interest is focused on attacking the commonly used graph models, which has led towards then using these insights to develop robust models that are resilient to these attacks.

Three key challenge problems involved in the modeling process include:

• Network alignment and merging: develop accurate and scalable methods for mapping of nodes across heterogeneous networks based on various associational and causal dependencies.
• Sub-graph detection and matching: develop robust and efficient algorithms for richly attributed networks to support detection and recognition of complex query patterns for networks.
• Attack and defense on graph data: develop imperceptible adversarial attack algorithms against existing graph models and develop resilient graph models that are robust against state-of-the-art attack methodologies.

The focus of this workshop is to gather together the researchers from all relevant fields to share their experience and opinions on addressing the three fundamental graph mining problems – “Connecting the dots”, “Finding a needle in a haystack”, and “Defending against attacks” in the context of adversarial activity analytics.

In addition, this workshop also aims to provide a forum for discussing research challenges and novel approaches in synthesizing realistic networks that those observed in the real-worlds. Numerous approaches have been proposed in the past for generating networks (e.g., exponential random graphs, stochastic Kronecker graphs). However, few research has been conducted on systematically injecting and embedding subtle signals (e.g., covert activities) to these “background” networks. In addition, new methods for generating synthetic data in other domains (e.g., Computer Vision) with deep generative models (e.g., Variational Autoencoders, Generative Adversarial Networks) have grown in prominence. Naturally, the question arises as to whether these new methods can be adapted to the graph domains and how they compare in capability to the current state-of-the-art. Besides the transaction-oriented networks that have been the main focus of this workshop, semantic networks (e.g., knowledge graphs) recently drawn significant attention in our research community. Thus, another important question is to extend the aforementioned graph computing capabilities to handle semantically rich networks to support the emerging research direction.

Topics of Interest

Including but not limited to:

Key Challenges:

• Network alignment and data integration from multiple heterogeneous domains
• Sub-graph detection and matching algorithms for large networks
• Attack and defense strategies on graph models (e.g., graph neural networks)
• Deep generative models for synthesizing realistic networks (e.g., static, dynamic, etc.)

Fundamentals:

• Limits of detectability and identifiability
• Fast and scalable graph mining algorithms
• Multilayer and multiplex network analytics
• Game theoretic approach on anticipating opponent intent and actions
• Novel evaluation metrics for graph mining and network analytics
• New methods and frontiers in spectral graph theory
• Identification of novel network datasets

Other related topics:

• Knowledge graph creation, analysis and application
• Graph neural networks (GNNs) and their application towards adversarial activity analytics
• Complex anomaly (e.g., group anomaly) detection and interpretation
• Clustering and ranking methods for composite networks
• Topological analysis (e.g., centrality and network motif analysis)
• Large-scale link prediction and recommendation algorithms
• Information diffusion and influence maximization
• (Interactive) visualization for big networks
• Semi-supervised learning, transductive inference, active learning, and transfer learning in the graph context

This workshop (co-located with the 2020 IEEE International Conference on Big Data) aims to bring together a cross-disciplinary audience of researchers from both academia and industry to share experience techniques, resources and best practices, and to exchange perspectives and future directions. We expect the workshop to develop a community of interested researchers and facilitate their future collaborations. A best paper will be selected and announced in our workshop based on the collective feedback from our reviewers.

Submission Instructions

Submissions to the workshop will be subject to a single-blind peer review process, with each submission reviewed by at least two program committee members in addition to an organizer. Accepted papers will be given either an oral or poster presentation slot, and will be published in the IEEE Big Data workshop proceedings.

Papers must be submitted in PDF format according to IEEE Computer Society Proceedings Manuscript Formatting Guidelines (Formatting Instructions, Templates) to fit within 10 pages (long papers), or 6 pages (short papers) including any diagrams, references and appendices. Submissions must be self-contained and in English. After uploading your submission, please check the copy stored on the site.

Submissions should be made using the Online Submission System provided by IEEE BigData.

Important Dates

• + Due date for full workshop papers submissionOctober 10, 2020
• + Notification of paper acceptance to authorsNovember 3, 2020
• + Camera-ready of accepted papersNovember 20, 2020
• + WorkshopDecember 10, 2020

Keynote Speakers

Prof. Leman Akoglu
CMU

Prof. William Hamilton
McGill University

Prof. Jingrui He
UIUC

Dr. Patrick Rubin-Delanch
University of Bristol

Accepted Papers

	Multi-Channel Entity Alignment via Name Uniqueness Estimation Miquette Orren, Patrick Mackey, Natalie Heller, and George Chin
	Robust and Scalable Entity Alignment in Big Data James Flamino, Christopher Abriola, Benjamin Zimmerman, Zhongheng Li, and Joel Douglas
	Who killed Lilly Kane? A case study in applying knowledge graphs to crime fiction Mariam Alaverdian, William Gilroy, Veronica Kirgios, Xia Li, Carolina Matuk, Daniel Mckenzie, Tachin Ruangkriengsin, Andrea Bertozzi, and Jeffrey Brantingham
	Evaluation of Alignment:Precision, Recall, Weighting and Limitations Joseph Cottam, Natalie Heller, Chrisopher Ebsch, Rahul Deshmukh, Patrick Mackey, and George Chin
	Using Graph Edit Distance for Noisy Subgraph Matching of Semantic Property Graphs Christopher Ebsch, Joseph Cottam, Natalie Heller, Rahul Deshmukh, and George Chin
	Data-Driven Template Discovery Using Graph Convolutional Neural Networks Mikel Joaristi, Sumit Purohit, Rahul Deshmukh, and Geroge Chin
	Semantic Guided Filtering Strategy for Best-effort Subgraph Matching in Knowledge Graphs Alexei Kopylov, Jiejun Xu, Kangyu Ni, Shane Roach, and Tsai-Ching Lu
	Fault-Tolerant Subgraph Matching on Aligned Networks Thomas Tu and Dominic Yang
	Static and Dynamic Social Network Models for the Analysis of Transshipment in Illegal Fishing Stefano Stamato and Andrew Park
	Graph Adversarial Attacks and Defense: An Empirical Study on Citation Graph Chau Pham, Vung Pham, and Tommy Dang
	Inexact Attributed Subgraph Matching Thomas Tu, Jacob Moorman, Dominic Yang, Qinyi Chen, and Andrea Bertozzi

Schedule

All times are US EST (Eastern Standard Time)

Program Committee

Program Co-chairs (Organizers)