Making everyday tech safer

NSF-backed UND project pushes to build smarter, longer-lasting defenses against cyberattacks at the “edge”

A heart monitor in a clinic. A sensor on a power line. A traffic camera at a busy intersection.

These everyday devices don’t sit behind the locked doors of a data center — they operate out in the world, at the “edge” of a network. And as edge devices become more common in healthcare, smart cities and energy systems, they’re also becoming more attractive targets for cyberattacks.

That’s why Jielun Zhang, assistant professor in the UND School of Electrical Engineering & Computer Science, is leading an NSF-supported project focused on building intrusion detection systems that work better in edge environments — where computing resources are limited, data is decentralized and threats evolve constantly.

“Our research focuses on developing sustainable and resilient cybersecurity systems … for distributed and edge computing environments,” Zhang said. “We aim to prevent failures that occur when traditional intrusion detection systems cannot keep up with new or evolving threats. These include zero-day attacks, data breaches, and disruptions caused by outdated or resource-intensive models.”

A smarter way to detect threats

Zhang’s approach treats edge security less like a single guard at a gate and more like a neighborhood watch.

Instead of sending sensitive raw data to a central hub, the research uses federated learning — allowing devices to train local models and share what strengthens collective defenses, without exposing private information. The goal is to help edge systems learn quickly while keeping sensitive data where it belongs.

Attackers change tactics — and many AI-based systems struggle to keep up because learning new patterns can come at the cost of forgetting old ones.

“Our goal is to create cybersecurity systems that not only adapt to new attacks but also remember what they’ve learned,” Zhang said, “preventing the kind of ‘forgetting’ that makes existing systems vulnerable over time.”

Ultimately, he added, “we want to make network defense systems that are both intelligent and enduring.”

The project also tackles a key tension in modern cybersecurity: protecting privacy while still enabling effective detection. “We aim to close the gap between privacy preservation and data utility by enabling secure, synthetic data generation that maintains usefulness for AI-driven monitoring without exposing sensitive information,” Zhang said.

Why this matters for North Dakota

The work lands at a moment when North Dakota’s digital footprint is expanding quickly — especially around high-performance computing and AI infrastructure.

Zhang sees UND’s research as a way to ensure that growth is matched with trustworthy security.

“Our work in sustainable, privacy-preserving intrusion detection can directly complement this growth,” he said, “by ensuring that the expansion of digital infrastructure is matched by strong, adaptive, and trustworthy security systems.”

He also points to the human side of that momentum: “Our educational and outreach efforts, from graduate research to K–12 STEM engagement, will help build the skilled workforce that will sustain this innovation across the state.”

Students at the center

“UND students are at the heart of this research,” Zhang said.

Students will help develop algorithms, build and deploy testbeds, and analyze results — and Zhang plans to integrate the latest findings into courses such as CSCI 471, CSCI 573 and CSCI 589, giving students direct exposure to emerging AI techniques and real cybersecurity applications.

“Through this experience,” Zhang said, “students gain not only the practical abilities that make them well prepared for cybersecurity careers in industry, but also the technical depth and creative thinking needed to become future leaders and innovators in the field.”

Written by Paige Prekker  //  UND College of Engineering & Mines