Short descriptors for the Community Meeting plenary talks can be found below
Day 1 Keynote
Date and Time: Thursday, August 22nd, 9:20AM-10:00AM
Speaker: Ilya Mironov, Meta
Bio: Ilya Mironov is a Sr. Staff Research Scientist, working on privacy-preserving machine learning at Meta. He previously worked at Google Brain (2015-2019) and Microsoft Research Silicon Valley Campus (2003-2014), where he contributed to early work on differential privacy. He holds a PhD in cryptography from Stanford. He is a member of the OpenDP Advisory Board.
Guest Talk #1: Actionable Underpinning Problems in DP
Date and Time: Thursday, August 22nd, 10:35AM – 11:20AM
Speaker(s): Gary Howarth (NIST), Christine Task (Knexus Research)
Descriptor: The question “What is epsilon?” is a perennial open problem in differential privacy. It has interdependencies with nearly all privacy research topics—privacy protection, utility, communication and implementation—and because it’s complex, it’s very difficult to make unambiguous progress to resolve it.
In this talk we propose four examples of a different type of open research problem: “Actionable Underpinning Problems” are essentially research linchpins. They have concretely defined contexts and objectives, so progress on them is less ambiguous. And, they also have a potentially big pay-off; they’re common points of uncertainty that underlie a much wider span of open research problems. In particular, progress on these would address uncertainties to help us better resolve the question of epsilon.
This talk draws from recent surveys of open problems in DP, and the presenters’ own experience supporting rapid research progress through challenges and benchmarks. For each underpinning problem we include toy illustrations and concrete project starters:
- tracking sources of error and randomness (and the difference between them)
- tracking algorithm dependency on data distribution properties
- tracking the knowledge complexity of privacy vulnerabilities
- tracking opportunities to streamline for usability
Industry Panel: Deploying DP in Practice
Date and Time: Thursday, August 22nd, 1:30PM – 2:30PM
Speaker(s): Yaw Etse (Capital One), Seyi Feyisetan (Amazon), James Honaker (Mozilla Anonym), Christina Ilvento (Apple), Rob Pisarczyk (Oblivious)
Descriptor: Last year there was an industry panel with a focus on understanding blockers to DP adoption and system building from researchers and product managers who worked in the DP space. The meta question perhaps was “Why aren’t there more DP systems in production?”
This year, we wanted to focus on sharing information of systems that have actually been built, and some of the stories highlighting the practical challenges in deploying those systems. The meta point might be “These are the barriers we’ve had to overcome.”
The two are related questions, but different perspectives. This is experience that feedback has said is extremely valuable and enlightening from attendees of the community meeting. This is also a way for the OpenDP core team to better prioritize and understand how tools could be used in practice.
Please note: this session will not be recorded
Guest Talk #2: Enhancing Open Data for Social Good: Leveraging capabilities of OpenDP to increase publication of socially relevant data
Date and Time: Friday, August 23rd, 9:10AM – 9:50AM
Speaker(s): Mayana Pereira (Microsoft)
Descriptor: This talk will explore the potential of OpenDP in enabling open data for social good. Focusing on the critical aspect of digital equity, we talk about open datasets created by Microsoft’s AI for Good team.
The 2020 broadband dataset, compiled to measure broadband usage at the zip code level across the United States, provides invaluable insights into the accessibility and utilization of broadband services, shedding light on the digital divide that exists in different regions in the U.S., especially the differences between urban and rural U.S.
Additionally, we will introduce the upcoming Digital Applications Index dataset, set to be published by the AI for Good team in 2024. This differentially private dataset offers a unique perspective by presenting metrics on the usage of digital applications at the zip code level. By leveraging OpenDP library, this dataset has the potential to empower researchers and policymakers to drive impactful change in areas such as economic development, education, and socio-economic analysis.
A key highlight of this talk is how now OpenDP functionalities, such as the additive noise mechanisms and the DP PCA recently integrated into OpenDP library, can democratize differentially private data publication. Publications of differentially private open data sets are usually performed by highly specialized teams in big tech companies such as Microsoft, Google, Meta, and LinkedIn. OpenDP tools have the potential of simplifying and speeding up data publication.
Day 2 Keynote: Government Activities to Advance Privacy Enhancing Technologies
Date and Time: Friday, August 23rd, 10:10AM-10:40AM
Speaker: Jeremy Epstein, White House Office of Science and Technology Policy (OSTP)
Bio: Jeremy Epstein is Assistant Director for Technologies and Privacy at the White House Office of Science and Technology Policy. Prior to joining OSTP, he led the National Science Foundation’s Secure and Trustworthy Cyberspace (SaTC) program, NSF’s flagship cybersecurity & privacy research funding program with over 1000 active research projects. He is past chair of the ACM’s US Technology Policy Committee (USTPC), and founder of Scholarships for Women Studying Information Security (SWSIS). He enjoys bicycling and chocolate, and is sorry he can’t be with you in person for this event.