Yiming Li

Yiming Li

Ph.D. Candidate, THU

Greetings and welcome to my homepage

Biography

I am currently a fourth-year Ph.D. candidate in Data Science and Information Technology from the Tsinghua-Berkeley Shenzhen Institute, Tsinghua University, advised by Professor Yong Jiang and Professor Shu-Tao Xia. Before that, I received my B.S. degree in Mathematics and Applied Mathematics (Yangming Innovation Class) from the Ningbo University in 2018, advised by Professor Lifeng Xi.

I was a research intern at Ant Security Lab (2021), working with Dr. Weifeng Qiu and Dr. Feng Xue; a research intern at Tencent AI Lab (2020, 2019), working with Dr. Baoyuan Wu and Dr. Zhifeng Li (supported by the Tencent Rhino-bird Elite Training Program).

I studied machine learning algorithms (especially tree-based ones) for their good interpretability and theoretical properties at the beginning of my post-graduate study. Currently, my research mainly focuses on AI security, especially backdoor learning, adversarial learning, and data privacy. My research is supported by the Tsinghua ‘Future Scholar’ Ph.D. Fellowship, and my Ph.D. dissertation topic is ‘Poisoning-based Backdoor Attacks in Computer Vision’.

I will graduate in June 2023 (expected). Currently, I am looking for both academic and industrial job opportunities (especially post-doc positions regarding AI security) starting from Summer 2023. Feel free to connect if there are suitable opportunities!

News

  • 12/2021: I will visit the Secure Learning Lab at UIUC, working with Professor Bo Li (start from 2022).
  • 12/2021: One paper is accepted by the AAAI 2022. Its codes have been released.
  • 11/2021: One paper is accepted by the Pattern Recognition.
  • 11/2021: One paper is accepted by the IEEE IoT Journal.
  • 09/2021: One paper is accepted by the Pattern Recognition.

Useful Resources

Github Repo about Backdoor Learning Resources

BackdoorBox (A Python toolbox for backdoor learning research.)

The ATT&CK Matrix of AI Security (The first technical report comprehensively covering different kinds of security threats in the full cycle of AI systems.)