Skip to main content

Tony (Lipeng) He

Any improvements made anywhere besides the bottleneck are an illusion. — Theory of Constraints

Tony He teaching an undergraduate computer science class

Teaching an undergraduate CS class

About

I'm a computer science PhD student at the University of Waterloo; part of Secure Systems Group (SSG), Cryptography, Security, and Privacy (CrySP) Lab, and the Cybersecurity and Privacy Institute (CPI).

I'm grateful to be advised by N. Asokan and Yaoliang Yu. I also worked with Jian Liu at Zhejiang University and did applied cryptography research at ABC Lab. Currently, my office is located in the William G. Davis Computer Research Centre, DC 3333.

I'm in pursuit of knowledge, experience, and the various other beautiful things life has to offer. I strive to live deliberately. Before research, I spent some years doing software engineering. In the limit of my life, I am also trying to be a pianist, writer, podcaster, designer, and entrepreneur[1].

[1] Retrograde Labs is a research-driven startup building the trust layer for AI. Our research aims not only to uncover failure modes and address fundamental bottlenecks, but to translate those insights into systems that work in production and withstand the test of the market and real customers. We build, grow, and scale products that turn research ideas into practical solutions for real-world workflows.

Research Interests

TL;DR: I study how adversarial robustness breaks when AI is deployed into real-world systems, and how to build defenses that survive those conditions.


My research focuses on Trustworthy Machine Learning, with an emphasis on the robustness of LLMs and the security & privacy of agentic AI systems. I study failures that appear when models are fine-tuned, approximated, compressed, connected to tools, exposed to external content, or placed in multi-agent, and other production ML pipeline settings. I am interested in issues such as data leakage, prompt injection, unsafe tool use, brittle safety behavior, and systems that cannot explain or audit what happened.

I develop effective and efficient adversarial attacks (e.g., automated red-teaming), as well as principled/scalable defenses (e.g., based on reinforcement learning, mechanistic interpretability), drawing on applied cryptography, theoretical machine learning, and systems security to characterize and mitigate emerging threats.

Research Vision

To ensure AI's transformative potential reaches as much of the society as possible in the most beneficially impactful way, with safety as the unlock that guarantees the world will benefit from AI.

Publications

* indicates equal contribution
Jiawen Zhang, Yangfan Hu, Kejia Chen, Lipeng He, Jiachen Ma, Jian Lou, Dan Li, Jian Liu, Xiaohu Yang, and Ruoxi Jia
Jiawen Zhang, Lipeng He, Kejia Chen, Jian Lou, Jian Liu, Xiaohu Yang, and Ruoxi Jia
Jiawen Zhang*, Kejia Chen*, Lipeng He*, Jian Lou, Dan Li, Zunlei Feng, Mingli Song, Jian Liu, Kui Ren, and Xiaohu Yang
Shoupeng Ren, Lipeng He, Tianyu Tu, Di Wu, Jian Liu, Kui Ren, and Chun Chen
Jiawen Zhang, Xinpeng Yang, Lipeng He, Kejia Chen, Wen-jie Lu, Yinghao Wang, Xiaoyang Hou, Jian Liu, Kui Ren and Xiaohu Yang
Di Wu, Shoupeng Ren, Yuman Bai, Lipeng He, Jian Liu, Wu Wen, Kui Ren, et al.

Preprints

Citations
Lipeng He, Yihan Wang, Jiawen Zhang, and N. Asokan
Under Submission
Vasisht Duddu, Mark Wang, Lipeng He, and Tim G. J. Rudner
Under Submission

SecStep: Guarding Agents Against Prompt Injection via Action Backtracking

Jiawen Zhang, Yechao Zhang, Lipeng He, Kejia Chen, Jiachen Ma, Yangfan Hu, Jian Lou, Jian Liu, Xiaohu Yang, and Tianwei Zhang
Under Submission
Jiawen Zhang, Kejia Chen, Jiachen Ma, Yangfan Hu, Lipeng He, Yechao Zhang, Jian Liu, Xiaohu Yang, Tianwei Zhang, and Ruoxi Jia
Under Submission
Lipeng He, Yihan Wang, Jiawen Zhang, and N. Asokan
Under Submission
Anudeep Das, Prach Chantasantitam, Gurjot Singh, Lipeng He, Mariia Ponomarenko, and Florian Kerschbaum
Under Submission

From Detection to Diagnosis: Lightweight Federated Prompt Learning for Interpretable Industrial Anomaly Analysis

Hao Pan, Xiaoli Zhao, Lipeng He, and Xiwu Shang
Under Submission

Token-by-Token Manipulation: Inference-Time Jailbreaking on Production LLMs via Autoregressive Harmful Guidance

Jiawen Zhang, Lipeng He, Kejia Chen, Jian Liu, Zunlei Feng, Mingli Song, Jian Lou, Dan Li, and Xiaohu Yang
Under Submission

Talks

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense

Academic Services

Program Committee Member

Conference

The Web Conference (WWW) 2027

USENIX Security Symposium (SEC) 2027

Artifact Evaluation Committee Member

Conference

USENIX Security Symposium (SEC) 2026

ACM Conference on Computer and Communications Security (CCS) 2025, 2026

Privacy Enhancing Technologies Symposium (PoPETs/PETS) 2026

🏆 USENIX SEC Noteworthy Reviewer, 2026

Reviewer

Journal

IEEE Transactions on Knowledge and Data Engineering (TKDE), 2026

IEEE Transactions on Dependable and Secure Computing (TDSC), 2026

Mentoring

Directed Reading Program (DRP) Mentor

University
AI Safety and Security Challenges in LLM-based Autonomous Agents (Spring 2026)

Women in Mathematics (WiM)

Funding

International Doctoral Student Award (IDSA)

University
CAD 5,100 per term

University of Waterloo

University of Waterloo Graduate Scholarship

University
CAD 4,000

University of Waterloo

AWS Startup Activate Credits (Portfolio)

Industry
USD 25,000

Amazon, Y Combinator

Lambda Research Grant Program

Industry
USD 5,000; Principal Investigator: N. Asokan

λ (Lambda) AI

David R. Cheriton Graduate Scholarship

University
CAD 20,000

University of Waterloo

International Master's Award of Excellence (IMAE)

University
CAD 2,500 per term

University of Waterloo

Teaching

University of Waterloo logo

University of Waterloo

University
  • CS 858 Trustworthy Machine Learning: a Systems Security Approach

    Course Staff (Expert)
  • CS 436 Networks and Distributed Computer Systems

    Teaching Assistant (TA)
  • CS 135 Designing Functional Programs

    Instructional Apprentice (IA)
  • CS 135 Designing Functional Programs

    Instructional Support Assistant (ISA)

Research Experience

Cryptography, Security, and Privacy (CrySP) Lab logo

University of Waterloo

Research, Part-time

Research Assistant (URA)

Cryptography, Security, and Privacy (CrySP) Lab

Zhejiang University logo

Zhejiang University

Research, Co-op

Research Assistant

ABC Lab, Institute of Cyberspace Research

Industry Experience

LinkedIn
Retrograde Labs logo

Retrograde Labs

Leadership

Member of Technical Staff

Accelerating scientific discovery and commercialization

BioRender logo

BioRender

SWE, Co-op

Full Stack Software Engineer

SaaS, Y Combinator W18

Toronto, ON

Safyre Labs logo

Safyre Labs

SWE, Co-op

Back End Software Engineer

E-Commerce Platform, Supply Chain

North York, ON

Bitbuy logo

Bitbuy

SWE, Co-op

Front End Software Engineer

Cryptocurrency Exchange, Part of Robinhood, TSX: WNDR

Toronto, ON

Education

University of Waterloo logo

University of Waterloo

PhD

Doctor of Philosophy

Computer Science

University of Waterloo logo

University of Waterloo

BMath

Honours Bachelor's Degree (Co-op)

Mathematics (Minor in Computing)

Nanyang Technological University logo

Nanyang Technological University

Undergrad

Exchange Student (GEM Trailblazer)

Mathematical Sciences

Newsletter

Podcast

New Article Everytime I Publish :)