Doctoral student Elizabeth Hou won an honorable mention for her presentation at the Michigan Student Symposium for Interdisciplinary Statistical Sciences (MSSISS). The MSSISS is organized by students in the Biostatistics, Electrical Engineering & Computer Science, Industrial & Operations Engineering, and Statistics & Survey Methodology departments with the purpose of encouraging communication across related fields of statistical sciences and promoting interdisciplinary research among graduate students and faculty.
Hou presented her paper, “Anomaly Detection in Partially Observed Traffic Networks.” The research concerns networks where messages are sent from node to node. It can be difficult to monitor these networks for security purposes (ensuring nothing is passing through the network that shouldn’t be there), because observers often/usually don’t have access to the messages themselves. To solve this problem, Hou suggests a system where you monitor the nodes rather than the messages.
“By watching the nodes, you can count how many messages are leaving and how many messages are entering,” Hou says. “And if you assume that your network has a certain type of structure, then you have a baseline to compare against. That way, it’s a lot easier to reconstruct your entire network and be able to detect if there’s any anomalous activity.”
While there are many applications of Hou’s research, her project was funded by the Department of Energywith the purpose of monitoring nuclear nonproliferation.
“In a nuclear field cycle, which is just a set of buildings, we attempt to track the patterns of shipping from building to building,” Hou says. “By using our anomaly detection methods, we can detect whether there’s proliferation or nonproliferation of nuclear materials.”
We can detect whether there’s proliferation or nonproliferation of nuclear materials.Elizabeth Hou, PhD candidate