Haochen Sun

Haochen Sun

Computer Science PhD Student

University of Waterloo

Biography

Haochen Sun is a PhD student at the Data Systems Group within the Cheriton School of Computer Science at the University of Waterloo, under the supervision of Prof. Xi He. His current research focuses on enhancing the security and privacy of machine learning and data management systems through the use of cryptographic tools and differential privacy techniques. He completed his undergraduate studies (diploma, transcript) in the Department of Mathematics and Department of Computer Science and Engineering at HKUST.

Interests
  • Machine Learning Security
  • Cryptography for Machine Learning
  • Differential Privacy
Education

  • PhD in Computer Science, 2022 - Present

    University of Waterloo

  • BSc in Data Science and Technology, and in Computer Science, 2018 - 2022

    The Hong Kong University of Science and Technology

Projects

*
All UWaterloo HKUST
Zero-knowledge Deep Learning
Specialized ZKP protocols for deep learning with CUDA implementations. First working ZKP scheme for 13B-size LLMs, and for training 10M-size neural networks.
Adversarial Example Tracing
Zero-shot tracing the origin of adversarial examples via watermarking.
Air Pollution Forecast with Deep Learning
Specialized deep-learning architecture for modelling the spatial-temporal distribution of air pollutants. Improving the accuracy of regional air-pollution forecast by 30%.

Selected Publications

Haochen Sun, Jason Li, Hongyang Zhang (2024). zkLLM: Zero Knowledge Proofs for Large Language Models. ACM Conference on Computer and Communications Security, CCS 2024, 14-18 October 2024, Salt Lake City, Utah, USA.

Preprint Code

Minhao Cheng, Rui Min, Haochen Sun, Pin-Yu Chen (2023). Identification of the Adversary from a Single Adversarial Example. International Conference on Machine Learning, ICML 2023, 23-29 July 2023, Honolulu, Hawaii, USA.

Cite URL

Haochen Sun, Jimmy C. H. Fung, Yiang Chen, Zhenning Li, Dehao Yuan, Wanying Chen, Xingcheng Lu (2022). Development of an LSTM broadcasting deep-learning framework for regional air pollution forecast improvement. Geoscientific Model Development.

Cite DOI URL

Contact

Please use your institutional email to contact me and identify yourself, as I cannot trust anonymous communications. Please use English, the official language of the university and the only language I can efficiently use for academic purposes. Limit our communications to email, Slack, MS Teams, and Zoom. Other communication channels are not endorsed by the university and are not installed on my work computers. Therefore, do not default to Google Meet or Tencent Meeting (we use Microsoft accounts) when scheduling a meeting with me, or ask for my WeChat or Telegram ID for collaboration or discussion. I appreciate your understanding.