OpenDP Platform for Differential Privacy

Differential privacy is a formal, mathematical conception of privacy preservation. An algorithm is differentially private when it injects a precisely calculated quantity of noise to any statistical query, masking the possible contribution of any one individual to the result. This provides a gold standard definition of privacy protection for data scientists who want to analyze data that contains personal information that must remain private. The open source project provides several basic building blocks that can be used by people involved with sensitive data, with implementations based on vetted and mature differential privacy research. It aims to connect theoretical solutions from the academic community with practical lessons learned from real-world deployments and to make differential privacy broadly accessible to future deployments. The core library is a native runtime that is built in Rust to be memory safe and fast. It can be used to safely build differentially private releases from native code running on various devices. It has a SQL data access layer that allows users to compose analysis graphs using the SQL language. The data access layer supports a wide variety of SQL database engines. To ease deployment, it includes a sample hosted service that shows users how to compose heterogeneous queries over the same dataset, fronted by a REST-based endpoint.