# Mengyuan (Marvin) Li

> Mengyuan (Marvin) Li is an Assistant Professor of Computer Science at the University of Southern California (USC), where he leads SEPT Lab (Security, Privacy, and Trust). His research covers verifiable AI systems, trusted execution environments (TEEs) and confidential computing, system and hardware security including side channels, and AI agent security.

This is the official homepage. The HTML pages are canonical; llms-full.txt contains their full text in Markdown.

Key facts:

- Position: Assistant Professor, University of Southern California (since 2024): Thomas Lord Department of Computer Science; Ming Hsieh Department of Electrical and Computer Engineering
- Lab: SEPT Lab (Security, Privacy, and Trust), https://septlab.com/
- Lab resource: verifiable.pro, https://verifiable.pro/, a public resource maintained by SEPT Lab with news, explainers, and guides on verifiable AI
- Recruiting: up to six PhD students for Fall 2027, together with Yue Zhao and Ruolin Li through a research cluster that spans their departments (co-advising possible, not required). Applications go through USC Viterbi Graduate Admission; applicants interested in working with him email mengyuanli@usc.edu with "[SEPT PhD]" in the subject line. Current USC MS and undergraduate students are welcome to work on research projects in SEPT Lab, and students at other universities can apply to collaborate on research (email subject "[SEPT Research]").
- Previously: postdoctoral researcher at MIT CSAIL (2022-2024, with Prof. Mengjia Yan); Ph.D. in Computer Science and Engineering, The Ohio State University (2022, advised by Prof. Yinqian Zhang)
- Recognition: CROSSLINE received the ACM CCS 2021 Best Paper Award Runner-Up; research on AMD SEV led to AMD security bulletins and CVEs (CVE-2020-12966, CVE-2021-26340, CVE-2021-46744, CVE-2023-20575)
- Identity: several researchers share the name Mengyuan Li. This one is at USC; Google Scholar user cgjqyuoAAAAJ; DBLP pid 157/4437-4
- Contact: mengyuanli@usc.edu; Thomas Lord Department of Computer Science, Ginsburg Hall (GCS), 1031 Downey Way, Los Angeles, CA 90089

## Main pages

- [Homepage](https://mengyuanli.com/): biography, research directions, recruiting, news, the publication list, teaching, and service
- [Join SEPT Lab](https://mengyuanli.com/#join): Fall 2027 PhD recruiting, research opportunities for USC MS and undergraduate students, and research collaboration for students at other universities
- [Publications](https://mengyuanli.com/#publications): published papers by year (papers from before 2024 are grouped together), then arXiv preprints, with DOI, PDF, code, and BibTeX links; each entry has a stable anchor such as #hollow-llm-sp26
- [Publications in BibTeX](https://mengyuanli.com/files/publications.bib): every entry of the publication list in one file

## Research topics

- [Verifiable AI](https://mengyuanli.com/verifiable-ai.html): zero-knowledge verification of LLM inference (Hollow-LLM Attack, IEEE S&P 2026), privacy-preserving model oversight (WAVE, ASPLOS 2026), and SEPT Lab's verifiable.pro resource
- [TEE and Confidential Computing](https://mengyuanli.com/tee-confidential-computing.html): TEE-based systems and their performance (confidential VMs, ORAM, confidential multi-GPU machine learning), attacks on AMD SEV and SEV-SNP (several led to AMD security bulletins and CVEs), defenses against ciphertext and speculative-execution side channels, and the security of serverless cloud platforms
- [AI Agent Security](https://mengyuanli.com/ai-agent-security.html): TEEs and runtime monitoring for observing and protecting agent execution; papers include WAVE, SCALE, CatchBench, Behavioral Skill Reconstruction, and infrastructure for verifiable agent memory, plus the open-source projects ClawGang and Muse MCP Adapter

## Teaching

- [CSCI 699: Understanding and Identifying Side-Channel Threats in Cloud and LLM Systems](https://mengyuanli.com/25fall_699.html): graduate seminar taught in Fall 2025 and Fall 2026; the page is the Fall 2025 syllabus
- [CSCI 699: Confidential Computing: Protecting Your Data on Cloud GPUs and CPUs](https://mengyuanli.com/course.html): graduate course taught in Spring 2025 and Spring 2026; the page is the Spring 2025 syllabus
- [CSCI 530: Security Systems](https://mengyuanli.com/#teaching): graduate course, upcoming in Spring 2027

## Lab and profiles

- [SEPT Lab](https://septlab.com/): lab website with current members, research, publications, and how to join
- [verifiable.pro](https://verifiable.pro/): SEPT Lab's public resource on verifiable AI
- [Google Scholar](https://scholar.google.com/citations?user=cgjqyuoAAAAJ)
- [DBLP](https://dblp.org/pid/157/4437-4.html)
- [USC Viterbi faculty profile](https://viterbi.usc.edu/directory/faculty/Li/Mengyuan)
- [SEPT Lab on GitHub](https://github.com/sept-usc)

## Optional

- [Full site text (llms-full.txt)](https://mengyuanli.com/llms-full.txt)
- [Sitemap](https://mengyuanli.com/sitemap.xml)
- [SoK: Understanding Design Choices and Pitfalls of Trusted Execution Environments (PDF)](https://mengyuanli.com/files/asiaccs_sok.pdf): ACM ASIACCS 2024
- [Hollow-LLM Attack (PDF)](https://mengyuanli.com/files/hollow-llm.pdf): IEEE S&P 2026
- [Hollow-LLM Attack BibTeX](https://mengyuanli.com/files/hollow-llm.bib)
- [CIPHERLEAKS project website](https://cipherleaks.com/)