Junda He

Papers

Total Citations

H-Index

About

Junda He is a rising researcher at the forefront of trustworthy and robust artificial intelligence, with a primary focus on the safety and reliability of sequential decision-making processes (SDPs). His work critically addresses vulnerabilities in deep learning-based systems used in high-stakes domains like autonomous driving, robotic control, and traffic management. He pioneered "Curiosity-Driven Testing," a novel methodology that intelligently explores SDPs to uncover hidden failures, earning 6 citations and setting a new standard for proactive system validation. Most notably, He is the lead author behind "BAFFLE," a groundbreaking study (with 8 combined citations across its 2022 and 2024 versions) that reveals a new class of security threat: backdoor attacks hidden within offline reinforcement learning datasets. This work exposes how malicious data providers can stealthily compromise learned policies, fundamentally challenging the trustworthiness of offline RL paradigms. By bridging the gap between cutting-edge AI and critical safety engineering, Junda He’s contributions are essential for building resilient autonomous systems. His research not only identifies critical vulnerabilities but also provides the foundational tools to detect and mitigate them, marking him as a key voice in the future of secure AI.