Weiyao Wang

Papers

Total Citations

H-Index

About

Weiyao Wang is a researcher at the forefront of machine learning security and privacy, with a particular focus on deep reinforcement learning. Her most cited work, "How You Act Tells a Lot: Privacy-Leaking Attack on Deep Reinforcement Learning" (2019, 38 citations), pioneered a novel attack vector demonstrating that an agent's behavioral patterns in reinforcement learning environments can be exploited to infer sensitive training data. This groundbreaking study revealed that even without direct access to model parameters or training datasets, adversaries could reconstruct private information simply by observing an agent's actions—a vulnerability previously overlooked in the field. A subsequent version of this work (14 citations) further refined the attack methodology, establishing Wang as a key voice in adversarial machine learning. Her contributions have significant implications for deploying RL systems in privacy-sensitive domains like healthcare, autonomous driving, and personalized recommendations. By exposing these behavioral privacy leaks, Wang has helped shape best practices for secure AI deployment, influencing both academic research and industry standards for privacy-preserving reinforcement learning. Her work continues to be cited as foundational reading for researchers exploring the intersection of privacy, security, and sequential decision-making in AI systems.