Hongbin Liu

Duke University, Delft University of Technology

Papers

Total Citations

H-Index

About

Hongbin Liu is a researcher specializing in 3D point cloud processing, adversarial machine learning robustness, and large-scale spatial data management. His most influential contribution, "PointGuard: Provably Robust 3D Point Cloud Classification" (2021), has garnered 61 citations and addresses a critical vulnerability in AI systems used for autonomous driving and robotic grasping — namely, their susceptibility to adversarial attacks that manipulate classifier predictions. By developing provable robustness guarantees for 3D point cloud classifiers, Liu has made meaningful strides in securing safety-critical perception systems against malicious perturbations. Beyond adversarial robustness, Liu has contributed to the foundational challenge of managing massive point cloud datasets at scale. His work on the nD PointCloud data structure (2018) and optimized space-filling curve approaches for n-dimensional window querying (2020) tackles the growing computational demands of robotics and virtual reality applications, where efficient storage and retrieval of large point clouds remain pressing bottlenecks. Together, these contributions span both the security and scalability dimensions of 3D data, positioning Liu as a researcher whose work is highly relevant to the practical deployment of modern perception and spatial computing systems.