OpenDP Progress and Outlook


As we draw to the close of an unusual year marked by the pandemic and the most difficult times it brought to us and our loved ones, the OpenDP team would like to extend our best wishes to you and your family this holiday season.  Since our Community Meeting in May, we have been hard at work on starting to materialize the vision that we discussed.  We would like to share with you our progress and our outlook for the coming year. 

What’s New 


We’re Hiring

We are currently hiring for our scientific staff (e.g. postdoctoral fellows).  If you or someone you know is interested in contributing technically to differential privacy system development, we will be very happy to hear from you.


The OpenDP Library

Based on the feedback received at the community meeting, we have defined the scope of our first product offering, which consists of the OpenDP library, a web-based application and integration with Dataverse repositories.

The OpenDP library is central to what we are doing.  We have begun development of the software library following the ideas described in the paper, "A Programming Framework for OpenDP."  Work has focused on a Rust implementation of the core framework concepts: measurements and transformations, privacy and stability relations with their associated distance measures, and chaining and composition operations.  We have also designed a schema for representing datasets to be processed by the library.  Our goal is to provide a model that's flexible enough for different algorithms and user-defined datatypes, yet still maintains good performance.  Finally, we've been working out the details of how library functionality will be exposed via a foreign function interface, and a strategy for writing higher-level language bindings.


Streamlined Design for A Web-Based Application

Through user experience studies with curators of large data repositories, we streamlined the design of a web-based application to enable users without coding experience to leverage the state-of-the-art OpenDP library and easily produce and manage differentially private releases in their use cases.  


Integration with Dataverse

The web-based application is designed to have integration with Dataverse repositories (, allowing owners and custodians of sensitive datasets on Dataverse to create and publish differentially private statistical releases.  Researchers can freely explore these releases while the access to the sensitive datasets remains restricted. 


Collaboration with Microsoft: SmartNoise

Our ongoing collaboration with Microsoft has produced and grown the SmartNoise product suite that is already powering sensitive data processing pipelines.  In expectation of the next magnitude of use cases across an increasing number of application domains in 2021, we are completing an extensive documentation project on SmartNoise, allowing us to scale up customer support.  


SmartNoise Early Adopter Acceleration Program

The SmartNoise Early Adopter Acceleration Program has just been announced by Microsoft.  We are excited to be a partner in helping participants fulfill their missions to benefit society. To apply for the program, please click here.


New Collaborations

OpenDP is developing partnerships with several other projects and organizations in the community.  For example, we have teamed up with the CrisisReady project to work on privacy-preserving data sharing for local decision-making in the face of crises such as COVID-19 and the California Wildfires.  We are in conversations with several other potential partners, which we hope to report on in the near future. 


Promoting Differential Privacy

Our Executive Committee members have been very active in promoting differential privacy to the public.  Videos and slides of some of their talks are available online.  Here are a few of the most recent ones: Keynote at TPDP, by Salil Vadhan, Red Hat Research Day: Privacy, by Merce Crosas and James Honaker, and MLSE 2020: Computing Systems, by Mercè Crosas.


The Domain

We are very thankful to Richard Attermeyer for the generous donation of the domain to us.  Thank you, Richard!


Coming Up Next

In the next three to six months, we have several major deliverable targets lined up back to back.   

The OpenDP library, along with its Python bindings, will be released with an initial set of algorithms, targeting basic summary statistics (mean, median, quantile, sum, count, variance), histograms, univariate regression, and covariance. The goal is to enable building simple applications, and to receive feedback on suitability of APIs.  The OpenDP Commons will be open to accept and vet the first community contributions.  The web-based application for accessing the library without coding experience will be in production with integration to Dataverse.  The documentation project will build on the SmartNoise work and cover our new releases. 

In addition to the SmartNoise Early Adopter Acceleration Program in our collaboration with Microsoft, we will launch an OpenDP Fellows Program for engineers or scholars in industry, government, or academia who wish to learn about and engage with the OpenDP software suite, bring this knowledge back to their own work or organization, while also contributing to OpenDP in the process. 

Our website will have a brand new look and enhanced community features.  Our Slack workspace will be open for use.  The next OpenDP community meeting is being planned for spring/summer 2021.


Join us and Stay Tuned

As soon as the OpenDP Commons is open for contribution, we will invite the community to build, review and use new DP algorithms, applications and integrations.  On our roadmap are more efforts to increase the breadth of the OpenDP offering.  So join our mailing list, drop us an email about your thoughts, ask questions on our Gitter, and create an issue or PR on our GitHub repositories anytime.  


We hope you stay safe and well this holiday season.  We look forward to working with you in the New Year to make great advancements in differential privacy.  



Annie Wu
Program Director for the Privacy Tools Project and OpenDP, IQSS
Harvard University