ECE CHIPS In: Efficiently Secure Computing
This article is the first part of a series highlighting BU ECE’s contributions to research and education under the umbrella of the CHIPS and Science Act of 2022, and our faculty’s collective investment in the national effort to build a robust and innovative semiconductor industry.
by A.J. Kleber
In today’s society we are constantly generating and sharing data, such as our aggregated fitness records, cell phone photo back-ups, cross-platform app preferences, and much more via any number of personal devices; most of which is sent to the cloud for processing. While these services make our lives easier in many ways, all this private data needs to be protected, to avoid malicious exploitation; and this is where some significant challenges come in that we, the end-users, generally don’t see.
Data encryption, implemented to protect user privacy, creates a dilemma: decrypting that data for processing, which is currently a common practice, leaves it vulnerable to attack. On the other hand, working with still-encrypted data is very resource-intensive, enough to challenge even the relatively enormous capacity of the cloud. With support from the Red Hat Collaboratory and the NSF, Professor Ajay Joshi and his research group are working to address this quandary: how can we improve computational efficiency without sacrificing security?
The answer may lie with Fully Homomorphic Encryption (FHE)–or rather, with the improvements to FHE that Professor Joshi’s team is working hard to develop. FHE is a method for processing data without decrypting it first; “homomorphic,” a Greek term meaning “same structure,” allows operations to be made on an analogous data set which preserves the relationship between elements of the original data. While it is a strong option as security goes, FHE has immense memory and computational requirements; it also generates a significant amount of “noise,” which must be cleared out in order to avoid corrupting the processing outcome. (For example, an email address could become corrupted with junk characters, in which case it would no longer work.) This cleanup process, termed “bootstrapping,” devours even more of the cloud’s already-taxed resources.
In order to transform FHE into a more efficient technique, Joshi and his students and multi-institutional collaborators have taken a variety of approaches, from designing novel architectures with built-in memory-saving techniques, to addressing computational congestion in “edge devices” –endpoints on the network, like your smartphone. Their endeavors have yielded four published papers so far this year, three at major computer architecture conferences (one at HPCA and two at MICRO 2023) and another in ”IEEE Transactions on VLSI Systems,” a top-tier journal. Three of these papers were first-authored by recent ECE graduates and former Joshi advisees Rashmi Agrawal (PhD’22) and Zahra Azad (PhD’22). All four papers describe significant improvement in the performance of FHE-based computation.
In their ongoing effort to improve the efficient process of protected data at every stage, Joshi’s group embodies a community-minded approach to research and technological development; not only working to protect the privacy of every individual in our interconnected society, but contributing new tools and techniques to a broad network of colleagues across different sectors, very much in the spirit of the Red Hat Collaboratory and societal engineering. Further complimentary projects and partnerships are under active discussion, which will no doubt continue to foster improvements to the evolving technological underpinnings of our everyday lives.
Professor Ajay Joshi joined the Department of Electrical and Computer Engineering in 2009. He is a Rafik B. Hariri Institute for Computing and Computational Science & Engineering Faculty Research Fellow and Affiliate, an affiliate of the BU Photonics Center, and serves as Associate Editor for IEEE Transactions on VLSI Systems. His honors include two Google Faculty Research Awards (2019 & 2018), a Best Paper Award at ASIACCS (2018) and HOST (2023), an NSF CAREER Award (2012), and many more. Professor Joshi’s research interests include computer architecture, hardware security and privacy, VLSI design, and silicon photonics.