Sareh Rowlands

Papers

Total Citations

H-Index

About

Sareh Rowlands is a researcher at the forefront of robotic perception and security, whose work critically examines the vulnerabilities of deep learning models in real-world autonomous systems. Her primary research areas span adversarial machine learning, object detection security, and human-robot interaction for assembly tasks. Rowlands made a significant contribution to the field with her 2023 paper, "Adversarial Detection: Attacking Object Detection in Real Time," which advanced beyond static image attacks to demonstrate how intelligent robots can be deceived in dynamic environments—a crucial step for deploying safe autonomous systems. Her 2024 follow-up, "A Human-in-the-Middle Attack Against Object Detection Systems," further exposed critical security gaps in embedded vision systems, highlighting the risks posed by increasingly powerful CPUs and GPUs in robotics. Earlier, Rowlands explored constructive human-robot collaboration in "Combining learning from demonstration and search algorithm for dynamic goal-directed assembly task planning" (2018), where she developed a novel method for robots to learn assembly plans from human demonstrations using CAD models. Though her citation counts are currently modest, her pioneering focus on real-time adversarial threats positions her as an emerging voice in robotic security, with work that is foundational for building resilient, trustworthy autonomous systems.