Big Graph Visualization

Dr. Song Zhang
szhang@cse.msstate.edu

Increasing amounts of data from graph & network sources are taxing analysts. Large graph data visualization has important applications in modern society including network security analysis, power grid management, social network mining, etc. Meanwhile, computer security logs for the basis of attack prevention and mitigation. However, the ever increasing size of the graph & log data poses a serious challenge to the limit of human visual perception. This project seeks to bridge the gap between the big graph & network log data and the user with scalable, customizable and uncertainty-aware visualization.

For the BigGraph Project, researchers at Mississippi State University and the Pacific Northwest National Laboratory (PNNL) will develop and deploy cutting-edge and pragmatic visualization technology to interactively navigate and explore very large data with a focus on semantic graphs modeled for or harvested from domain applications such as cyber network alert, Web and sensors. The proposed research will target visualizing graph datasets of unprecedented size and complexity as deemed necessary by the funding agency. Much of the computation R&D will be conducted on Puma—a newly established high-performance computer located at PNNL and Shadow 2 located at MSU. The project will use benchmark datasets harvested from the public domain to demonstrate its scalability and efficiency. PNNL researchers will be responsible for handling technology transfer from R&D to end users at the agency. This project will be supported by a team of PNNL and MSU senior researchers and graduate students who have backgrounds in visualization, computation, applied mathematics and data science.

We will build a visual analytics system prototype that supports users to visually explore and analyze big graphs such as a network flow dataset stored in a triple store database system on Puma. The ultimate goal is to visually explore big graphs with tens of billions of edges, which is unprecedented in terms of scope and implications.

For the NetTriage Project, researchers at MSU working with PNNL will develop effective methods to quickly and accurately process large amounts of streaming network alerts, either from the graph systems under development or extant alert systems. The work will allow triage and analysis of network traffic and varying scales to better support cybersecurity analysts.