Connected Vehicle Data Privacy: Lessons Learned

Abstract

Data privacy has been a key focus area for USDOT as it supports the development and deployment of connected vehicle technology. The Connected Vehicle Data Privacy (CVDP) Project led to many useful transportation-related privacy insights, tools, and stakeholder engagements. In 2013 and 2014, Oak Ridge National Laboratory (ORNL) scientists analyzed connected vehicle field test data as specified by current standards in an attempt to re-identify the driver. The outcome of that work was a Privacy Protection Algorithm (PPA) for large databases containing geolocation traces generated by connected vehicles. The need to incorporate map data into the algorithm was the critical observation. Map information allows privacy protections to be measurable, less heuristic, and more adaptable to variations in road network structure. This algorithm was used to process and release a portion of the Ann Arbor Safety Pilot data into the public domain. Other findings included the ability to use highly distinguishable vehicle length and width measurements, physics-based, and timing-based attacks to link trip segments across pseudonym changes. Mitigations have been integrated into SAE standards (J2945) for vehicle dimensions. Timing-based attacks must be mitigated through improvements to On-Board Equipment (OBE) clock synchronization. Mitigation of physics-based attacks (decreasing location and time fidelity) could impact safety applications, so they are not recommended. Since 2015, ORNL has developed a Privacy Protection Module for streaming data processed by the Operational Data Environment (ODE), an ASN.1 codec for the ODE, a Privacy Sensitivity Model, and a Graphical User Interface (GUI) and other technical updates for the PPA. ORNL also continued to evaluate the privacy protections provided by connected vehicle technology and promote practices that support transportation data privacy among community stakeholders. Through this project, USDOT has conducted extensive analysis and vetting of connected vehicle data privacy concerns. The tools created have been released as open source software. Our techniques have been shared with a broad audience that includes state transportation departments, privacy companies, policy makers and agencies, standards bodies, academic researchers, and federal agencies. This report outlines some high-level lessons learned, recommendations for technology deployers and data curators, and a list of future research directions.