Sarus framework enables privacy-preserving perception fusion for autonomous vehicles

Researchers have developed Sarus, a novel framework designed to enable privacy-preserving fusion of perception data from multiple autonomous vehicle vendors. This system utilizes homomorphic encryption to allow a central fusion server to aggregate detection outputs without accessing sensitive individual vendor data or proprietary model behaviors. Experiments on the KITTI dataset, using detectors like YOLOv8 and PointPillars, demonstrate that Sarus can effectively enhance scene-level coverage by combining complementary detections, particularly in scenarios where individual sensor modalities might falter. AI

IMPACT Enables secure collaboration between different autonomous vehicle systems, potentially accelerating multi-vendor development and deployment.

RANK_REASON Academic paper detailing a novel privacy-preserving framework for autonomous vehicle perception fusion. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• autonomous vehicles
• homomorphic encryption
• KITTI dataset
• PointPillars
• PV-RCNN
• Sarus
• YOLOv8

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →