Session Track: Graph + AI in Production
Session Time:
Session description
Efficiently querying an incomplete cyber knowledge graph consisting of vulnerabilities, weaknesses, attack patterns, mitigations, and attack entities is a challenging problem. We will present a hybrid approach by exploiting the graph structure (or topology) and semantic information (using Query2Box) in the knowledge graph. Our approach will address shortcomings of incomplete information, memory footprint, and retrieval efficiency. By building a GraphRAG interface, we aim to enable cyber defenders to build threat intelligence and make informed decisions. We will present an empirical evaluation with real-world examples.
Chief Scientist and Group Leader, Pacific Northwest National Laboratory
Dr. Halappanavar is a chief data scientist at PNNL, where he serves as the group lead of the Data Science and Machine Intelligence group. He also holds a joint appointment as adjunct faculty in computer science at Washington State University in Pullman. His research has spanned multiple technical foci and includes combinatorial scientific computing, parallel graph algorithms, artificial intelligence and machine learning, and the application of graph theory and game theory to solve problems in application domains, such as scientific computing, power grids, cybersecurity, and life sciences. He has authored over 160 technical publications for peer-reviewed journals, conferences, and workshops.
Deputy CISO, Pacific Northwest National Laboratory
Joseph Aguayo builds cybersecurity, data science, and AI applications for safety-critical and highly-regulated industries. Prior to joining PNNL, Joseph served in cybersecurity leadership roles in the banking / financial industry and at the US Department of Defense. As a computer scientist, he has held academic appointments at NC State University and the University of Maryland.
Postdoctoral Researcher, Pacific Northwest National Laboratory
I am a Postdoctoral researcher at Pacific Northwest National Laboratory (PNNL). I received my Ph.D. From department of Computer Science at Purdue University. My research interests lie broadly in Graph-based Machine Learning, Natural Language Processing, and cybersecurity. During my PhD, I designed scalable algorithms for Graph Neural Networks. I have also worked on classifying and analyzing Software Vulnerabilities into Weaknesses and Attack Patterns using large language models. Google Scholar: https://scholar.google.com/citations?user=kTAFl2yYe6QC&hl=en
University of Texas at El Paso
I'm a Ph.D. student in Computer Science at the University of Texas at El Paso, where I also completed my M.S. degree. My research focuses on AI-driven educational tools and cybersecurity, with recent work on knowledge graphs and retrieval-augmented models. I interned at Pacific Northwest National Laboratory for the Summer 2024 and 2025, and had previously served as a faculty member at Shahjalal University of Science and Technology in Bangladesh. My research has been presented at workshops and conferences like KDD, CIKM, and IEEE Big Data, and I enjoy mentoring, teaching, and exploring real-world applications of machine learning.
Data Engineer, Pacific Northwest National Laboratory
Samantha Silva is PNNL's Cybersecurity Data Engineer for the Advanced Introspection & Analytics team. She has experience in software development, AI and cybersecurity. She is responsible for designing and implementing secure, scalable data pipelines that process and automate large scale information flows. Prior to becoming a Cybersecurity Data Engineer, Samantha was a Software Engineer Intern at NASA Marshall Space for Flight Center. She also did numerous of research work in Cybersecurity and AI at her university. Samantha received both a Bachelor of Science in Computer Science and a Master of Science in Software Engineering with a concentration on Secure Cyber Systems from the University of Texas at El Paso, Go Miners!