Shai Ben-David, Ian Goldberg named 2023 ACM Fellows

Wednesday, January 24, 2024

Professors Shai Ben-David and Ian Goldberg have been named 2023 ACM Fellows. The Association for Computing Machinery is recognizing Professor Ben-David for his contributions to and research leadership in machine learning theory, and Professor Goldberg for his contributions to the development and deployment of privacy enhancing technologies.

The Association for Computing Machinery is the world’s largest educational and scientific computing society, uniting computing educators, researchers and professionals to inspire dialogue, share resources and address the field’s challenges. ACM Fellowships are conferred to the top one percent of the association’s members. This prestigious recognition indicates outstanding accomplishments in computing and information technology and exemplary service to ACM and the computing community.

This year the Association for Computing Machinery named 68 members across the globe as ACM Fellows for their wide-ranging and fundamental contributions in disciplines running the gamut of the computing field, including algorithm design, computer graphics, cybersecurity, energy-efficient computing, mobile computing, software analytics, and web search, to name a few. Fellows are nominated by their peers, with nominations reviewed by a distinguished selection committee.

“Congratulations to Shai and Ian on their being named ACM Fellows,” said Raouf Boutaba, Professor and Director of the Cheriton School of Computer Science. “Their recognitions are well deserved. Both are exceptional researchers whose pioneering and sustained contributions, mentorship of students, and service to the community have shaped and informed the direction of their respective fields within computer science.”

With Professors Ben-David and Goldberg’s nomination, 10 faculty members at the Cheriton School of Computer Science have been named ACM fellows. Previous recipients of this prestigious honour are Professors N. Asokan, Srinivasan Keshav, Ihab Ilyas, Ming Li, Jimmy Lin, J. Ian Munro, M. Tamer Özsu and Frank Tompa.

About Professor Shai Ben-David

photo of Professor Shai Ben-David
Shai Ben-David is a Professor and University Research Chair at the Cheriton School of Computer Science as well as a Canada CIFAR Artificial Intelligence Chair

He holds a PhD in mathematics from the Hebrew University in Jerusalem, and has held postdoctoral positions at the University of Toronto in both the mathematics and computer science departments. He was a professor of computer science at the Technion in Haifa, Israel, before joining the Cheriton School of Computer Science.

Professor Ben-David’s research spans learning theory, computational complexity and mathematical logic. A prolific researcher, he has published more than 130 papers that have been cited collectively more than 23,000 times with an h-index of 51 as of January 2024 according to Google Scholar. His publications have received best paper awards at five conferences — at ICASSP 2005, COLT 2006, COLT 2011, NeurIPS 2018 and ALT 2023.
 
With Shai Shalev-Shwartz, he co-authored Understanding Machine Learning: From Theory to Algorithms, a textbook published in 2014 by Cambridge University Press that introduces machine learning and the mathematical foundations on which it is based. This publication is both a standard textbook as well as a research reference.

His lectures on machine learning, logic and computation on YouTube have been viewed thousands of times by students and colleagues alike, with his introductory lecture on machine learning garnering more than 100,000 views to date.

RESEARCH CONTRIBUTIONS

Foundational Research in Machine Learning
Professor Ben-David began investigating machine learning in the early 1990s, well before it reached its current ubiquity. His early works uncovered key issues that later became central to machine learning. 

Computational hardness of machine learning tasks
Professor Ben-David established some of the first results on computational hardness of learning. These include the first hardness of approximation result for learning neural networks in “Hardness results for neural network approximation problems,” a EuroCOLT 1999 paper with Peter Bartlett, as well as NP-hardness of learning for natural concept classes in “On the difficulty of approximately maximizing agreements,” a paper with Nadav Eiron and Philip Long presented at COLT 2000.

Change-detection in data streams
Professor Ben-David was among the first to recognize the importance of addressing non-stationary data generation in machine learning. His key publications in this area are “Learning changing concepts by exploiting the structure of change,” a COLT 1996 paper with Peter Bartlett and Sanjeev Kulkarni, “Detecting change in data streams,” a VLDB 2004 paper with Daniel Kifer and Johannes Gehrke, and “Nonparametric change detection in 2D random sensor field,” a paper with Ting He and Lang Tong that also received a best paper award at ICASP 2005.

Multi-class and real valued learning
Professor Ben-David played a key role in developing combinatorial dimensions that characterize learnability beyond binary classification. His key publications include “Scale-sensitive dimensions, uniform convergence and learnability,” a 1997 paper with Noga Alon, Nicolò Cesa-Bianchi and David Haussler, “Characterizations of learnability for classes of {O, …, n}-valued functions,” a COLT 1992 paper with Nicolò Cesa-Bianchi and Philip Long, and “Multiclass learnability and the ERM principle,” a paper with Amit Daniely, Sivan Sabato and Shai Shalev-Shwartz that also received the best paper award at COLT 2011.

Initiating Novel Research Directions

Domain adaptation
Professor Ben-David’s publication “Analysis of representations for domain adaptation,” a NIPS 2006 paper with John Blitzer, Koby Crammer and Fernando Pereira, propelled research on what is now recognized as key to machine learning applications. This paper continues to inspire domain adaptation algorithms and is a major theoretical benchmark.

Methodical analysis of clustering paradigms
Professor Ben-David’s NIPS 2008 paper “Measures of clustering quality: A working set of axioms for clustering” with Margareta Ackerman initiated a long line of research, developing a theory for taxonomizing clustering methods according to their input-output behaviour. This paper alerts users to the crucial need for thoughtfully choosing the clustering method for a given task. Follow-up papers by Professor Ben-David and his coauthors introduce tools for that important and under-investigated issue.

Learning with computable learners
Professor Ben-David’s ALT 2020 paper “On learnability with computable learners,” with Sushant Agarwal, Nivasini Ananthakrishnan, Tosca Lechner and Ruth Urner, initiated research on learnability under computability requirements. Surprisingly, much of common learnability theory collapses under this natural condition. His follow-up paper “On computable online learning,” with recent master’s graduate Niki Hasrati, which expanded on these results, won the ALT 2023 best paper award.

Independence of set theory results in computer science
Professor Ben-David’s 1991 paper “On the independence of P versus NP” with S. Halevi proved equivalence between the independence from a strong axiomatic theory of P vs NP and having almost-polynomial-time SAT solvers.

And his invited STOC 2021 paper “Learnability can be independent of set theory” with Pavel Hrubes, Shay Moran, Amir Shpilka and Amir Yehudayoff showed that learnability of some basic tasks cannot be decided by common mathematics. This is the first result in machine learning establishing independence from Zermelo–Fraenkel set theory with the axiom of choice.

Proving Unexpected Limitations of Popular Algorithmic Paradigms

Randomization in online learning
Professor Ben-David’s 1994 paper “On the power of randomization in on-line algorithms” with Allan Borodin, Richard Karp, Gábor Tardos and Avi Wigderson showed that the then-commonly held belief that randomization may help the competitive ratio performance of online algorithms is severely limited if not essentially incorrect.

Limitations of learning via embeddings in Euclidean half spaces
Professor Ben-David’s COLT 2001 paper “Limitations of learning via embeddings in Euclidean half-spaces” with Nadav Eiron and Hans Ulrich Simon proved inherent shortcomings of the popular support vector machines paradigm.

Theoretical analysis of the stability method in clustering
Professor Ben-David’s award-winning COLT 2006 paper, “A sober look at clustering stability” with Ulrike von Luxburg and Dávid Pál, indicated fallacies in common usage of stability as a tool for choosing clustering parameters.

SERVICE CONTRIBUTIONS
Professor Ben-David has also served the computer science and machine learning community in a variety of administrative capacities. He served as president of the Association for Computational Learning from 2009 to 2012, as the action editor for the journal Machine Learning, on various National Science Foundation and European Union funding agency panels, and as the program committee chair for all theoretical machine learning conferences — the European Conference on Computational Learning Theory in 1997, the Conference on Learning Theory in 1999, and the International Conference on Algorithmic Learning Theory in 2004. He also serves regularly as the senior area chair for the largest annual machine learning conferences — the Conference on Neural Information Processing Systems, the International Conference on Machine Learning, and the Conference on Artificial Intelligence and Statistics.
 
Professor Ben-David has also organized workshops on emerging research directions. These include the 2016 Dagstuhl seminar on Foundations of Unsupervised Learning in Germany, the 2016 Lorentz workshop on the Theoretical Foundations of Learning from Easy Data in The Netherlands, the Simons Institute’s cluster on Interpretable Machine Learning in 2020 and 2022 in Berkeley, California, and the 2023 IDEAL Workshop Machine Learning Interpretability and Logic in Chicago.

About Professor Ian Goldberg

photo of Professor Ian Goldberg
Ian Goldberg is a Professor at the Cheriton School of Computer Science and the Canada Research Chair in Privacy Enhancing Technologies, a prestigious national research position he has held since 2019. He has a master’s and a PhD degree in computer science, both from the University of California, Berkeley.

Professor Goldberg’s ACM Fellowship is among his growing list of professional recognitions, awards and accolades. These include being named an IEEE Senior Member in 2023, receiving the USENIX Security Test of Time Award in 2019 and the Caspar Bowden Award for Outstanding Research in Privacy Enhancing Technologies in 2018. In addition, he was named an ACM Distinguished Member in 2017, received a Pioneer Award from the Electronic Frontier Foundation in 2011, an Outstanding Young Computer Science Researcher Award from the Canadian Association of Computer Science in 2011, and an Early Researcher Award from CS-Can | Info-Can in 2010. 

Professor Goldberg is the only person to date whose papers have been recognized three times — in 2013, 2018, and 2023 — with the Andreas Pfitzmann Best Student Paper Award at the Privacy Enhancing Technologies Symposium. This award recognizes the best paper appearing at that year’s symposium where a student has co-authored and presented the paper. 

RESEARCH CONTRIBUTIONS
Professor Goldberg is a leader in privacy enhancing technologies — what are known as PETs — a field in which he has been working for more than a quarter of a century. He has created, developed and deployed many innovations that have substantially and meaningfully improved privacy as well as access online against a backdrop of increased censorship, surveillance by governments, and data misuse and abuse by corporations.

His work appears regularly not only in PoPETs — the Proceedings on Privacy Enhancing Technologies known previously as PETS or the Privacy Enhancing Technologies Symposium — the top research venue dedicated to this field (22 papers over 2007–2024), but also at top-tier venues in the broader computer security and privacy community. These include the ACM Conference on Computer and Communications Security (12 papers over 2009–2023), the USENIX Security Symposium (10 papers over 1996–2023), the IEEE Symposium on Security and Privacy (five papers over 2007–2015), and the Network and Distributed System Security Symposium (two papers over 2013–2018). 

All told, he has published 111 peer-reviewed papers to date, of which 37 have been cited more than a hundred times, and two more than a thousand. Recently, he was ranked the top researcher in Canada assessed by the number of papers published at the top security and privacy venues, and the top researcher in the world for the same measure in PoPETs/PETS.

Professor Goldberg is known internationally for his research on a variety of privacy enhancing technologies, including secure and private messaging, privacy-preserving communications networks, private computation, censorship resistance, and more. His contributions range from mathematical foundations to protocol and system design to implementation and deployment. The privacy enhancing technologies he and his students have developed and deployed allow for personal autonomy online for citizens of restrictive regimes and in Western democracies alike.

His work on Off-the-Record Messaging brought widespread encrypted messaging to Internet users. His work on privacy-preserving communications networks, such as his contributions to the widely used, open-source Tor network, helps millions of people around the world protect themselves and their online communications from both censorship and surveillance. As many people live in countries where the Internet is censored, Professor Goldberg’s research on censorship resistance technologies allow them to access the free and open Internet. In this area, his research on Telex (2011) and Slitheen (2016–2021) allow people in censored countries to access websites with the help of a cooperating Internet service provider outside the censoring country. To the overseeing authority, it appears the user is accessing an innocuous website — videos of cats — while the Internet service provider secretly redirects the user’s traffic to the desired censored website. Even his team’s proof-of-concept deployment of Telex saw real censored users being helped by the technology. More recently, his open source Lox software (2023), which received one of three Andreas Pfitzmann Best Student Paper Award at PETS 2023 with his recent master’s student Lindsey Tulloch, similarly allows people in censored countries to access the Tor network and from there, the open Internet. This work is being enhanced by The Tor Project to deploy to Tor’s millions of daily users.

Professor Goldberg is also well known for his work on private computation, where a computation involving private data can be performed without the parties performing that computation being able to see that private data. Private computation prevents the kinds of large-scale data breaches seen almost daily, as companies would be able to perform the computations they desire without being able to access the underlying unencrypted data. If they do not have access to the data themselves, it cannot then be stolen from them. Professor Goldberg has worked on a variety of aspects of private computation, including private information retrieval, trusted execution environments, and multiparty computation, including the multiparty digital signature scheme FROST. FROST is being standardized in the Crypto Forum Research Group of the Internet Research Task Force and incorporated into a number of blockchain products.

Professor Goldberg’s success is attributable largely to building systems that solve the whole problem for real-world users, not just an academically interesting subset. This guiding principle enables his work to be adopted in a way that scales to serve everyone. 

Contributions of Record

A number of Professor Goldberg’s innovations have seen wide impact and deployment, a sample of which are detailed below.

From 2004–16, he was the lead researcher and developer of Off-the-Record Messaging, a protocol to provide secure and private messaging to users of existing instant messaging networks. Hundreds of thousands of people around the world used OTR to protect their online communications. His desktop-focused OTR protocol was later adapted to mobile environments as the Signal protocol. More than a billion people worldwide now use the Signal protocol as part of Signal, WhatsApp, and Facebook Messenger mobile apps.

With his work on privacy-preserving and censorship-resistant communications networks — from his PhD thesis to his award-winning PoPETs paper in 2023 — he has seen many of his ideas being deployed, first as the commercial Freedom Network in the early 2000s where he was Chief Scientist, then into the open-source Tor network where he was Chair of the Board of Directors for 10 years. Millions of people every day use Tor to protect their privacy, and directly benefit from Professor Goldberg’s research.

Professor Goldberg’s work with University of London’s Alex Davidson and others in a paper titled “Privacy Pass: Bypassing Internet Challenges Anonymously” aims to reduce the prevalence of CAPTCHAs, an online test to differentiate people from bots. It does so by having one party that can verify a user that then issues the user with tokens, which can then be redeemed at websites to bypass CAPTCHAs. The tokens are issued blindly, using a cryptographic technique that ensures that even the issuing party cannot recognize the issued tokens when they are subsequently redeemed. This work is being standardized at the Crypto Forum Research Group, and is being rolled out as Apple’s Private Access Token feature in iOS 16 and macOS Ventura.

SERVICE CONTRIBUTIONS
Professor Goldberg has held many leadership positions in the privacy enhancing technologies and broader computer security and privacy research communities. He was Program Co-Chair for PETS (2008, 2009), Program Chair for the USENIX Security Symposium (2010), and Program Co-Chair for Financial Cryptography and Data Security (2019). He was the General Chair for PETS 2011. 

He serves on the Privacy Enhancing Technologies Advisory Board (2007 to present), where he is responsible for monitoring their financial position, and for maintaining the article submission system and the website software infrastructure. He also implemented the move to self-publishing for the 2022 Proceedings on Privacy Enhancing Technologies. He has served on 36 program committees, including PETS/PoPETs, ACM CCS, USENIX Security Symposium, IEEE Symposium on Security and Privacy, ACM AsiaCCS, ACM WiSec, and many others.

Professor Goldberg served as the Chair of the Board of Directors of The Tor Project for the first 10 years of the organization’s existence (2007–2016), and currently serves on the Board of Waterloo’s Cybersecurity and Privacy Institute (2017–present), a cross-campus interdisciplinary institute bringing together researchers studying all aspects of cybersecurity and privacy, from cryptography and data science to human and societal aspects of security and privacy to law and policy.