Moh Kashani working to safeguard medical devices
Author: Zach Clemens
Author: Zach Clemens
Medical sensor technology has advanced leaps and bounds in the 21st century, and with that comes the need to secure those devices against attacks so those using sensors can get accurate information they can trust.
Think of a continuous glucose monitor (CGM), a wearable sensor used by individuals who have type 1 diabetes. The device continuously monitors blood sugar levels, letting the user always know their numbers. But what if those numbers weren’t accurate, or even worse, they were intentionally changed? It could be fatal for a person with diabetes.
Computer engineering graduate Moh Kashani is focusing his research on safeguarding these medical devices. CGMs and insulin pumps can be attacked by nefarious actors and must be secured so users can trust their information is accurate.
“What can we do to ensure these devices are authentic and accurate?” says Kashani. “We started looking at the wireless physical layer signal — can we authenticate the wireless signal from these devices to know they are real and accurate?”
This is the question Kashani is working on now, alongside Professor and Palmer Department Chair in Electrical and Computer Engineering Ashfaq Khokhar and Associate Professor Sang Kim, in collaboration with Professor Farid Nait-Abdesselam from Université Paris Cité.
They explored this idea of authentication and had the opportunity to record signals with specialized equipment from multiple sensors inside an anechoic chamber located on campus.
“After that, we developed a machine learning model that can authenticate using the physical layer signal. But at the next step, we realized the model was insufficient because adversarial attacks can happen on machine learning models,” Kashani says. “The adversarial attack can change the signal just slightly, which would cause the model to fail.”
Kashani will be working on safeguarding these machine learning models further at the National Science Foundation headquarters in Virginia. He was accepted into the Networking Technology and Systems Early-Career Investigators (NeTS-ECI) Workshop. He was invited to attend based on using an anomaly-based approach to make resiliency in deep learning-based Radio Frequency Fingerprinting.
“What we want to do is develop a new model that can identify an anomaly — an attack — and remove that anomaly, therefore only accepting valid signals,” says Kashani.
NeTS-ECI allows early-career researchers to engage in discussions with both leaders in the field and their peers to share perspectives and explore new research frontiers.
Kashani will also be able to showcase his research interests, build a collaborator network, kick-start collaborations, and benefit from professional development activities.