Chengze Du

I am currently pursuing my bachelor's degree at the School of Cyberspace Security, Beijing University of Posts and Telecommunications(BUPT), from 2021 to the present.

I' m passionate about travel, marathons. In academic terms, my research interests include but not limited: Network Tomography, Deep Learning, Differential Privacy, Artificial intelligence of things.

Email  /  CV  /  Google Scholar  /  Gitee

profile photo

Research

Some works are highlighted.

2025's Works
SecureNT alternate SecureNT
Beyond Single-Text Analysis: A Holistic Approach to Chinese Financial Sentiment
Mingfei Zhang, Chengze Du
International Conference on Advanced Computational Intelligence   (ICACI 2025)
[📁Code] [📜Paper] [😀arXiv]

This paper introduces a comprehensive framework that addresses the Single-Text sentiment analysis's challenge by context-aware architectures, and multi-factor sentiment fusion, offering a more robust solution for Chinese financial sentiment analysis.

SecureNT alternate SecureNT
GuidedLatent: Defending VAEs against Membership Inference Attacks via Distribution-Guided Privacy
Chengze Du, Guangzhen Yao, Jibin Shi, Ying Zhang, Renda Han
International Joint Conference on Neural Networks   (IJCNN 2025)
[📁Code] [📜Paper] [😀arXiv]

This paper introduces a controlled distribution mechanism that dynamically adjusts latent representations based on semantic similarities, coupled with a two-phase training strategy that gradually incorporates privacy constraints.
This work was done during the Internship in Zhipu.

2024's Works
SecureNT alternate SecureNT
SecureNT: Smart Topology Obfuscation for Privacy-Aware Network Monitoring
Chengze Du, Jibin Shi, Hui Xu, Guangzhen Yao
International Conference on Intelligent Computing   (ICIC 2025)
[📁Code] [📜Paper] [😀arXiv]   Oral

This paper introduces a novel privacy-preserving framework, which provides efficient topology protection while maintaining the utility of measurements for authorized network monitoring.

clean-usnob Identification of Path Congestion Status from End-to-End Measurements Using Deep Spatial-Temporal Learning
Chengze Du, Zhiwei Yu, Xiangyu Wang
Computer Communications
[📁Code] [📜Paper] [😀arXiv]

This work introduces the concept of Additive Congestion Status to address these challenges effectively. Using a framework that combines Adversarial Autoencoders (AAE) with Long Short-Term Memory (LSTM) networks, this approach robustly categorizes and quantifies the Additive Congestion Status.

Education & Experience

BUPT Beijing University of Posts and Telecommunications
B.E. in Cyberspace Security
Sep 2021 - Jun 2025
  • Thesis Title: Secure training methods for large language models based on differential privacy
  • Supervisor: Jinguo Bi
Internship Zhipu AI
Intern in AI Department
Oct 2024 - Jan 2025

Contributions

Reviewing

Misc

  Interests: Passionate about running, with personal bests of 21:21 for 5km and 1:45:28 for the half marathon


Last updated: 2024-12-12. And thanks this website's source code.