Donghang Duan

Donghang Duan

Robust FL / LLM Safety

Publications

You can also find my articles on my Google Scholar profile.

Look Twice before You Leap: A Rational Agent Framework for Localized Adversarial Anonymization

Published in arXiv preprint, 2025

This paper addresses the privacy paradox in LLM-based anonymization by proposing RLAA, a fully localized and training-free framework that introduces a rationality gatekeeper to prevent utility collapse.

Recommended citation: Donghang Duan, Xu Zheng. (2025). "Look Twice before You Leap: A Rational Agent Framework for Localized Adversarial Anonymization." arXiv preprint arXiv:2512.06713.
Download Paper

Exploiting Parasitic Dependency for Free-Rider Elimination in Blockchain-based Federated Learning

Published in 2025 IEEE International Performance, Computing, and Communications Conference (IPCCC), 2025

An incentive mechanism uncovering free-riders in blockchain-based federated learning.

Recommended citation: Donghang Duan, Xu Zheng, Yifu Zheng, Chong Mu, Ruozhou Wang, Ke Yan. (2025). "Exploiting Parasitic Dependency for Free-Rider Elimination in Blockchain-based Federated Learning." 2025 IEEE IPCCC.
Download Paper