CCNI Workshop

The 8th Workshop on Graph Techniques for Adversarial Activity Analytics (GTA³ 2024)

October 25, 2024 | Boise, Idaho, USA

For past GTA3 workshops, please visit our archive page.

Workshop Summary

Graphs are powerful analytic tools for modeling adversarial activities across a wide range of domains and applications. Examples include identifying and responding to cybersecurity systems' threats and vulnerabilities, strengthening critical infrastructure's resilience and robustness, and combating covert illicit activities that span various domains like finance, communication, and transportation. With the rapid development of generative AI, the lifecycle and throughput of adversarial activities, such as generating attacks or synthesizing deceptive signals, have accelerated significantly. For instance, a malicious actor can generate a large number of malware variants to flood defense systems or create agents to disseminate misleading signals, obscuring their activities. Consequently, there is a pressing need for novel and effective technology to autonomously handle these adversarial activities and keep pace with the evolving threats. The purpose of this workshop is to provide a forum to discuss emerging research problems and novel approaches in graph analysis for modeling adversarial activities in the age of generative AI.

Theme and Topics

Adversarial activities are often covert and embedded across multiple domains, making them generally undetectable and unrecognizable when viewed in isolation. They only become apparent when analyzed jointly across these domains. Therefore, a main research focus in modeling adversarial activities is developing techniques to fuse information from different networks into a unified view for comprehensive analysis. Equally important is the detection and matching of indicative patterns to recognize underlying adversarial activities within activity networks or graphs. Additionally, sophisticated adversarial actions may involve attempts to cover their tracks by attacking and altering networks, which has led to interest in attacking graph machine learning models. This, in turn, drives the development of robust models resilient to such attacks. The goal of the workshop is to address three fundamental problems in graph-based adversarial activity analytics: "Connecting the Dots," "Finding a Needle in a Haystack," and "Defending Against Attacks." In particular, with the advent of generative AI, we aim to augment state-of-the-art techniques and tackle these challenges in this new era.

In addition to analyzing activity networks, knowledge graphs play an important role in adversarial activity analysis. They provide the benefits of contextualizing and correlating information based on facts, enhancing the ability to detect and understand complex adversarial patterns. For instance, in the context of cybersecurity, knowledge graphs can help attribute attacks to specific threat actors by linking patterns of behavior, tools, and techniques to known profiles of adversaries. Similarly, in financial systems, knowledge graphs can be used to detect fraudulent activities by connecting transactions, accounts, and entities to identify unusual patterns indicative of fraud.

Topics of Interest

Including but not limited to:

  • Graph alignment and data integration from multiple heterogeneous domains
  • Subgraph detection and discovery algorithms for large networks
  • Attack and defense strategies on graph models (e.g., GNNs)
  • Generative models for synthesizing realistic networks (e.g., static, dynamic, etc.)
  • Representation learning on graphs
  • Multilayer and multiplex networks
  • High-performance graph computing
  • Explainability of graph models (e.g., GNNs)
  • Limits of detectability and identifiability
  • Link prediction and recommendation
  • Information diffusion and influence maximization
  • New methods for clustering and ranking on graphs
  • Novel datasets and evaluation metrics for network analytics
  • Knowledge graph creation, mining, and applications
  • Knowledge graph completion and reasoning
  • Complex anomaly detection and interpretation
  • Summarization and visualization of large networks
  • Interactive graph search and exploration
  • Topological analysis (e.g., motif analysis) on large graphs
  • Game-theoretic approach for adversarial modeling on graphs
  • Graph-based semi-supervised learning, active learning, and transfer learning
  • Frontiers of graph machine learning for adversarial activities analyticss
  • Large language models for adversarial activities.
  • Weakly supervised and self-supervised anomaly detection.
  • Fairness, explainability and privacy in adversarial activities analytics.
  • Benchmarks for adversarial activities analytics.

Submission Instructions

Submissions to the workshop will be subject to a double-blind peer review process, with each submission reviewed by at least two program committee members and an organizer. Submissions will be evaluated based on their relevance to the workshop, scientific novelty, and technical quality. Accepted papers will be given a presentation slot in the workshop.

Papers must be submitted in PDF format according to the ACM CIKM template. Submissions can vary in length from 4 to 8 pages, plus additional pages for references (not counted towards the page limit). Note that there is no distinction between long and short papers; authors may decide on the appropriate length of their paper.


Submissions should be made using the Online EasyChair submission system

Important Dates (tentative)

  • + Submission due date:8/16/2024
  • + Notification due date:8/30/2024
  • + Camera-ready of accepted papers9/20/2024
  • + Workshop date:10/25/2024

Keynote Speakers

We plan to invite speakers who are experts in relevant research areas. Please see our past workshops for more information: https://ccni.hrl.com/workshop/archive


Prof. Arijit Khan
Aalborg University


Prof. Jiliang Tang
Michigan State University


Dr. Yinglong Xia
Meta


Dr. Mahantesh Halappanavar
PNNL

Accepted Papers

Narrative Analysis of True Crime Podcasts With Knowledge Graph-Augmented Large Language Models

James Chapman, Xinyi Leng, Jason Liang, Jack Mauro, Xu Wang, Andrea Bertozzi, Junyuan Lin, Bohan Chen, Chenchen Ye, Temple Daniel and P. Jeffrey Brantingham

Rethinking Temporal Graph Transformers for Outlier Detection

Kay Liu, Jiahao Ding and Mohamadali Torkamani

Robustness of Knowledge Graph Embedding Approaches under Non-targeted Adversarial Attacks

Arnab Sharma, Sourabh Kapoor, Michael Roeder, Caglar Demir and Axel-Cyrille Ngonga Ngomo

Schedule

Times are listed in local time, MST

08:55 | Opening Remarks
09:00 | Keynote 1: User-friendly Explanations for Graph Neural Networks - Prof. Arijit Khan
09:40 | Keynote 2: Learning on Graphs: What is Next? - Prof. Jiliang Tang
10:20 | Narrative Analysis of True Crime Podcasts with Knowledge Graph-Augmented Large Language Models
10:35 | Coffee Break
11:00 | Keynote 3: Graph Analytics on Exascale Systems - Dr. Mahantesh Halappanavar
11:40 | Keynote 4: A Tale of Two Cities in Recommendation: Graph and LLM - Dr. Yinglong Xia
12:20 | Rethinking Temporal Graph Transformers for Outlier Detection
12:35 | Robustness of Knowledge Graph Embedding Approaches under Non-targeted Adversarial Attacks
12:50 | Closing

Program Committee

  • Bo Ni (Vanderbilt University, USA)
  • Dana Warmsley (HRL Laboratories, USA)
  • Dawei Zhou (Virginia Tech, USA)
  • Fan H Hung (HRL Laboratories, USA)
  • Harry Shomer (Michigan State University, USA)
  • Jian Kang (University of Rochester, USA)
  • Jun Wu (Michigan State University, USA)
  • Kang-Yu Ni (HRL Laboratories, USA)
  • Kevin Martin (HRL Laboratories, USA)
  • Lihui Liu (Wayne State University, USA)
  • Michael Warren (HRL Laboratories, USA)
  • Nigel Stepp (HRL Laboratories, USA)
  • Qinghai Zhou (Meta, USA)
  • Rodolfo Valiente Romero (HRL Laboratories, USA)
  • Sam Johnson (HRL Laboratories, USA)
  • Tameez Latib (HRL Laboratories, USA)
  • Xin Wang (Vanderbilt University, USA)
  • Yu Wang (University of Oregon, USA)

Program Chairs (Organizers)