Data Mapping at a Billion Dollar Self-Driving Vehicle Startup

At Cruise, Privacy Engineering faces a large scale data mapping challenge: maintaining visibility on petabytes of sensitive data collected daily–location, imagery, and copious metadata. If that weren’t enough, data mapping must be done in the context of a rapidly changing product as we launch self-driving cars to the public. This talk describes how we did it.

In more detail, we summarize the background research that informed our approach to data mapping. Then, we cover key aspects of the technical implementation: a custom web application, high accuracy detectors for Cruise-specific sensitive data, and a validation stack for machine-assisted label verification. Finally, we present the technical privacy risk mitigations we have built on top of data mapping. These pieces all fit together into a foundational capability for Cruise Privacy Engineering, greatly aiding our ability to understand the flow of sensitive data in a way that is not possible without automation.