Privacy-Aware Sharing of Raw Spatial Sensor Data for Cooperative Perception

TL;DR

This work argues that raw spatial sensor data sharing for cooperative perception promises substantial gains but introduces new privacy and IP leakage risks that impede adoption by automakers. It introduces SHARP, a privacy-aware framework combining location obfuscation via real-time novel-view rendering with an open, standardized sensor data stack and a governance/cost-model roadmap to incentivize cross-vendor participation. Feasibility analyses indicate that raw data can reveal vehicle pose, underscoring the need for effective obfuscation, while initial results show real-time novel-view synthesis as a viable privacy mechanism. The paper highlights open questions in vision foundation models, data fidelity trade-offs, and regulatory and standards development, aiming to catalyze multi-stakeholder discussions and international collaboration toward practical raw-data cooperative perception.

Abstract

Cooperative perception between vehicles is poised to offer robust and reliable scene understanding. Recently, we are witnessing experimental systems research building testbeds that share raw spatial sensor data for cooperative perception. While there has been a marked improvement in accuracies and is the natural way forward, we take a moment to consider the problems with such an approach for eventual adoption by automakers. In this paper, we first argue that new forms of privacy concerns arise and discourage stakeholders to share raw sensor data. Next, we present SHARP, a research framework to minimize privacy leakage and drive stakeholders towards the ambitious goal of raw data based cooperative perception. Finally, we discuss open questions for networked systems, mobile computing, perception researchers, industry and government in realizing our proposed framework.

Figures (6)

• Figure 1: (a) shows the privacy leakage due to raw data streaming from blue car to the green car; (b) abstracts typical sensor architectures and shows raw data types
• Figure 2: SHARP's approach to hide true location: use fast 3D understanding and rendering to constantly spoof realistic, raw data from a random view point.
• Figure 3: Preliminary simulation on the OPV2V scenes.
• Figure 4: SHARP's approach to minimize IP leakage: Open sensor configurations, open low-level software and standardizing hardware agonstic raw data as output from sensor hardware. Blue blocks are proprietary. We schedule (b) to create shareable raw data, else we schedule a fully proprietary stack (a).
• Figure 5: A demo on Town06 of OPV2V dataset, where contributor pretends it is at location 2.