Wenhan Chang

PhD Researcher in AI Safety & Privacy

Zhongnan University of Economics and Law | Wuhan, China

GitHub Scholar Zhihu

About Me

I received my B.Eng. degree in 2022 and M.Eng. degree in 2025 from China University of Geosciences, Wuhan. I am currently pursuing a Ph.D. degree at School of Information Engineering, Zhongnan University of Economics and Law.

My research interests include security and privacy preserving in deep learning, with a focus on LLM safety alignment, poisoning attacks and defenses, and machine unlearning.

2025 - 2029 (Expected)

Ph.D. in Modern Technology Management

Zhongnan University of Economics and Law

Microeconomics, Econometrics, Business Intelligence

2022 - 2025

M.S. in Electronic Information

China University of Geosciences

Machine Learning, Information Security, Matrix Analysis

2018 - 2022

B.E. in Computer Science and Technology

China University of Geosciences

C++ Programming, Embedded Systems, Computer Network

Publications

As First/Corresponding Author

2026

Zero-shot Class Unlearning via Layer-wise Relevance Analysis and Neuronal Path Perturbation

W Chang, T Zhu, P Xiong*, Y Wu, F Guan, W Zhou

IEEE Transactions on Dependable and Secure Computing (CCF-A)

2026

Unreal Thinking: Chain-of-Thought Hijacking via Two-stage Backdoor

W Chang, T Zhu, P Xiong, F Guan, W Zhou

arXiv preprint arXiv:2604.09235

2025

Chain-of-Lure: A Synthetic Narrative-Driven Approach to Compromise Large Language Models

W Chang, T Zhu*, Y Zhao, S Song, P Xiong, W Zhou, Y Li

arXiv preprint arXiv:2505.17519

2025

From Thinking to Output: Chain-of-Thought and Text Generation Characteristics in Reasoning Language Models

J Liu, Z Xu, Y Fang, Y Chen, Z Ying, W Chang*

KSEM 2025 (CCF-C)

2024

Class Machine Unlearning for Complex Data via Concepts Inference and Data Poisoning

W Chang, T Zhu*, H Xu, W Liu, W Zhou

arXiv preprint arXiv:2405.15662

2024

Gradient-based Defense Methods for Data Leakage in Vertical Federated Learning

W Chang, T Zhu*

Computers & Security, 139, 103744 (CCF-B)

As Co-author

2025

Large Language Models Merging for Enhancing the Link Stealing Attack on Graph Neural Networks

F Guan, T Zhu*, W Chang, W Ren, W Zhou

IEEE Transactions on Dependable and Secure Computing (CCF-A)

2025

TeleAI at SemEval-2025 Task 8: Advancing Table Reasoning Framework with Large Language Models

S Xiong, M Li, D Wang, Y Zhao, J Zhang, C Pan, H He, X Li, W Chang, et al.

SemEval 2025 (ACL Workshop)

2025

T2R-BENCH: A Benchmark for Real World Table-to-Report Task

J Zhang, C Pan, S Xiong, K Wei, Y Zhao, X Li, J Peng, X Gu, J Yang, W Chang, et al.

EMNLP 2025 (CCF-B)

2025

TableReasoner: Advancing Table Reasoning Framework with Large Language Models

S Xiong, D Wang, Y Zhao, J Zhang, C Pan, H He, X Li, W Chang, Z He, et al.

arXiv preprint arXiv:2507.08046

2024

Generative Adversarial Networks Unlearning

H Sun, T Zhu*, W Chang, W Zhou

IEEE Transactions on Dependable and Secure Computing (CCF-A)

2024

A Two-stage Model Extraction Attack on GANs with a Small Collected Dataset

H Sun, T Zhu*, W Chang, W Zhou

Computers & Security, 137, 103634 (CCF-B)

2020

Model Poisoning Defense on Federated Learning: A Validation Based Approach

Y Wang, T Zhu*, W Chang, S Shen, W Ren

NSS 2020 (EI)

* indicates corresponding author. Full list on Google Scholar

Wenhan Chang

About Me

Ph.D. in Modern Technology Management

M.S. in Electronic Information

B.E. in Computer Science and Technology

Publications

As First/Corresponding Author

Zero-shot Class Unlearning via Layer-wise Relevance Analysis and Neuronal Path Perturbation

Unreal Thinking: Chain-of-Thought Hijacking via Two-stage Backdoor

Chain-of-Lure: A Synthetic Narrative-Driven Approach to Compromise Large Language Models

From Thinking to Output: Chain-of-Thought and Text Generation Characteristics in Reasoning Language Models

Class Machine Unlearning for Complex Data via Concepts Inference and Data Poisoning

Gradient-based Defense Methods for Data Leakage in Vertical Federated Learning

As Co-author

Large Language Models Merging for Enhancing the Link Stealing Attack on Graph Neural Networks

TeleAI at SemEval-2025 Task 8: Advancing Table Reasoning Framework with Large Language Models

T2R-BENCH: A Benchmark for Real World Table-to-Report Task

TableReasoner: Advancing Table Reasoning Framework with Large Language Models

Generative Adversarial Networks Unlearning

A Two-stage Model Extraction Attack on GANs with a Small Collected Dataset

Model Poisoning Defense on Federated Learning: A Validation Based Approach

Research Interests

LLM Safety Alignment

Privacy Preserving

Poisoning Attack & Defense

Machine Unlearning