skip to main content

CS Faculty Fletcher and Solomonik Earn NSF CAREER Awards

7/8/2020 1:50:37 PM Laura Schmitt, Illinois CS

Two Illinois CS assistant professors have received CAREER awards from the National Science Foundation—the most prestigious awards in support of early-career faculty who have the potential to serve as academic role models in research and education and to lead advances in the mission of their department or organization.

Christopher Fletcher received $468,000 in funding to develop techniques that prevent cloud-based processors from leaking sensitive data like passwords or online transactions through a malicious attack.

Christopher Fletcher
Christopher Fletcher

According to Fletcher, these attacks can occur when a hacker’s program runs on the same processor simultaneously as an individual’s transaction or computation. The processors leave behind traces of data that can be picked up by the attacker’s software.

“The first order of business is to better understand the hardware we run our programs on, from a security perspective,” Fletcher said. “Processors are extremely complex and their implementation is, for the most part, opaque to users.  This prevents us from being able to write safe programs.”

In order to stay a step ahead of potential attacks, Fletcher proposes obtaining certain processor specifications from manufacturers like Intel and AMD. If he had these specifications that completely laid out where the leakage sources might be, he could close off the channels and have the upper hand, said Fletcher.

His approach involves designing a Distinguishability Set Architecture (DSA) for the processors, which is meant to be a peer to existing Instruction Set Architectures. According to Fletcher, the DSA specifies under which conditions each instruction reveals secret information. With a DSA, programmers or compilers can tune sensitive programs to avoid leaking secrets.

The educational component of Fletcher’s award includes scaling up a successful high school outreach effort, where he currently teaches students to create applications like chatbots, and implementing similar instruction as tutorials at conferences.

See NSF's award abstract.

Edgar Solomonik received $500,000 in funding to develop new parallel software and algorithms for tensor computations, which are a class of challenging calculations characterized by a large number of parameters or unknown variables. Tensors provide a mathematical framework for solving complex and data-intensive problems in physics, chemistry, and bioinformatics.

Edgar Solomonik
Edgar Solomonik

The high-level tensor library that Solomonik produces will accelerate analysis of large graphs, approximation of multidimensional datasets by tensor decompositions, and simulation of quantum systems—the latter being a major focus for Solomonik and his research group.

“We will extend an open-source library that we have on GitHub,” Solomonik said. “We’ll be able to handle new tensor kernels more efficiently than before and have much larger capabilities than we do now.”

Solomonik envisions that he and other researchers can deploy the tensor-based techniques on massively parallel computing systems, making new innovations in computational science possible.

The education component of Solomonik’s grant includes the development of new online learning modules for understanding tensor methods and computations.

“Tensors are becoming increasingly of interest in a variety of domains but there are very few courses or materials out there for learning about them in a systematic way,” he said. “We will develop a new course and methodology of tensor thinking, which aims to express algorithms and applications in terms of common tensor kernels.”

Solomonik also aims to produce a system that predicts the performance of a parallel tensor code without actually running it in parallel.

See NSF's award abstract.