AV

Khaled Khasawneh receives NSF Awards in collaboration with University of California

In This Story

People Mentioned in This Story
Body

, assistant professor in the department of Electrical and Computer Engineering and the director of the Computer Architecture, Machine Learning, and Security (CAMLsec) Lab, has been awarded two NSF grants. These grants are in collaboration with University of California (UC) Davis and UC Riverside, and total $2.4M.

“These are timely awards to allow my group to continue pursuing cutting edge research in machine learning security, cloud security, and hardware security fields,” says Khasawneh. “Special thanks to NSF for their funding. I appreciate the efforts of my collaborators, the endless support from our department, and the efforts of the talented students in my group.”

Award 1:

Cloud computing paradigms have emerged as a major facility to store and process massive amounts of data produced by various business units, public organizations, Internet-of-Things, and cyber-physical systems. The cloud scheduler is the component responsible for deciding which computer a cloud application should run. The current design of cloud schedulers only focuses on meeting the performance requirements of submitted applications without security considerations.

This project, in collaboration with professor Houman Homayoun’s lLab at UC Davis, examines how cloud schedulers can be exploited by attackers to facilitate targeted micro-architectural attacks in cloud environments. The project also explores novel approaches to defend against targeted micro-architectural attacks in the cloud.

Award 2:

Advances in Deep Neural Networks (DNN) have enabled a wide range of promising applications. However, DNNs are vulnerable to Adversarial Machine Learning attacks, with potentially dangerous outcomes, such as mistaking a stop sign for a speed limit sign.

This project, in collaboration with professors Nael Abu-Ghazaleh and Samet Oymak at UC Riverside, will explore the use of approximate computing to improve the robustness of DNNs against adversarial attacks. Approximate computing is a design paradigm that trades results precision for simpler hardware.