In This Story
Jianli Pan is working at the leading edge of emerging technologies, with recent funded projects covering such topics as analyzing cybersecurity metadata traffic; improving automated and intelligent security for internet of things (IoT) edge systems; and using deep reinforcement learning (DRL) techniques for cloud resource allocation for future IoT applications.
An associate professor in the Department of Information Sciences and Technology (IST), Pan explained that early in his career, he received good advice about where to focus his research. “The technology landscape is changing every five years and my PhD advisor—Raj Jain at Washington University in St. Louis—who is a very smart man and famous in our field, told me, ‘Always do something that the world needs. Invest your time and your energies in that direction and it will always pay off.’”
This advice led Pan to focus on what is broadly called cyber threat intelligence, focused on applying AI and machine learning technologies perform metadata traffic and system analysis. This will help make smart, efficient security systems for everyday devices—like sensors, cameras, and other connected gadgets—by spotting possible threats, weak points, and attacks right where the data is created, instead of relying on the cloud. He said researchers like himself now have the ability to apply novel technologies to solve long-standing problems, like making sense of massive amounts of data. “We can train language models to model those suspicious or potentially malicious behaviors from attackers and synthesize and gain useful insights out of this data.”
Pan is affiliated with the college’s biggest research center, the Center for Secure Information System (CSIS). Collaborating with colleagues in this group, he has received funding from the National Science Foundation (NSF) and the National Security Agency (NSA) to work on related projects.
Pan and his PhD student are currently exploring multi-modal methods of generative AI variants for some specialized tasks, like analyzing those huge amounts of data, particularly network traffic and miscellaneous system data. Pan explained that when identifying threats on the Internet, the key information is frequently buried in a massive amount of irrelevant data. This new approach will make finding that easier.
One challenge he notes is that security management is very siloed—diverse devices with diverse capabilities, cripple effective interactions or integrations among or between the silos. “Amazon provides services for Amazon devices and their apps. Google provides services for their own vertical products. There are barely any effective interactions or integrations among the different silos and that’s bad because hacks can come horizontally from the customers’ device side. And then the vulnerability in one point of weakness can be propagated across devices and there are huge consequences.”
Pan said that’s why he’s working on a different paradigm, using data-driven, machine-learning AI technology to make things more human-friendly. Humans are error-prone and can’t monitor systems around-the-clock. So, Pan said, “We need to build novel methods to liberate the human from these tedious, repetitive tasks.” Current NSA funding has him looking at exactly those kinds of tools.
Pan’s work reflects a growing shift toward AI-driven security systems that can keep pace with modern threats. By rethinking how we analyze data and manage risk, he’s helping lay the groundwork for a more resilient and adaptive cybersecurity landscape.