PIVOT Catalyzing the Community Around Collection and Sharing of Automotive Research Datasets 1

### Overview In this presentation at the ASRG Secure Our Streets conference, David Balenson (associate director and interim director of the Networking and Cybersecurity Division at USC Information Sciences Institute) introduces the PIVOT project. PIVOT is a collaborative effort involving USC ISI, Colorado State University, the University of Memphis, and industry partner Geotab. Funded under the US National Science Foundation (NSF) Computer and Information Science and Engineering (CISE) Community Research Infrastructure (CCRI) program, the grant began in October 2022 with plans to run through 2025 or 2026. Rather than conducting primary research itself, PIVOT focuses on developing a community platform to curate, collect, and share automotive and heavy-duty vehicle datasets and tools for vehicle cybersecurity, intelligent transportation, and smart/connected communities.

### Key Topics The project is structured around five core pillars: - **Platform:** The core interactive software/hardware infrastructure hosted at the University of Memphis (mirrored at CSU), incorporating a web portal, a catalog based on CooNDA (from the NSF Clavata/SEARCH projects), an Object Store using MinIO, and an authentication/naming shim. - **Data:** Three distinct sources of data, spanning passenger cars and heavy trucks. - **Tools:** Software handling multiple layers of the vehicular data pyramid, from physical layer oscilloscope traces and raw CAN/PDU frames to vehicle signals, fleet aggregations, and high-level analytical tools. - **Services:** Access control, cataloging, Institutional Review Board (IRB) support, and data privacy services. - **Community:** Outreach via workshops, webinars, academic/industry partnerships, and educational student challenges.

### Technical Details PIVOT categorizes its datasets into three primary types: 1. **Community Datasets:** Prior research datasets made visible through PIVOT's portal, such as Oak Ridge National Laboratory's ROAD dataset, Korea University HCRL datasets, Bosch's sinAn, and CSU heavy truck datasets. 2. **Geotab Data:** Sourced via the PIVOT Spindle program (a fleet of ~10 researcher vehicles equipped with Geotab GO9 telematics loggers sharing detailed trip, engine fault, fuel, and speed data via MyGeotab) and Geotab's Altitude analytics platform (providing aggregated transportation and stop analytics across over 4 million vehicles, specifically offering datasets for Denver, CO and Shelby County, TN). 3. **Custom CAN Logger 4:** An open-source, low-cost recording unit equipped with GPS, Bluetooth, Wi-Fi, CAN, and Automotive Ethernet recording capabilities intended for crowdsourcing vehicular data across diverse models.

Balenson also reviews outcomes from the April 2024 PIVOT Community Workshop, highlighting Mike Westra's keynote on topics beyond CAN (such as Adaptive AUTOSAR, Automotive Ethernet, and EV charging infrastructure). Workshop breakout sessions emphasized challenges in vehicle data time synchronization across heterogeneous buses, edge filtering, and privacy-preserving techniques.

### Takeaways Automotive network datasets have historically been fragmented, ad hoc, and challenging to acquire due to architectural variations across vehicle makes and privacy concerns. PIVOT addresses these barriers by establishing standardized community infrastructure, engaging students through competitions (CyberAuto, CyberTruck, CyberTractor, and CyberBoat), and offering data anonymization tools. During the Q&A, Balenson addressed privacy handling (such as Geotab geofencing and Privacy-Enhancing Technologies) and encouraged researchers to contribute datasets, adopt PIVOT infrastructure, and collaborate on future proposals.