Research
My research interests center on security boundaries for modern computing systems, particularly where autonomous software, operating systems, and storage hardware meet.
AI Agent Security
AI agents can operate software tools and manage files with broad permissions. I study the security implications of privileged agents and the risks of unauthorized modification or deletion of sensitive data.
Operating Systems and Kernel Security
I am interested in Linux access-control and enforcement mechanisms, including AppArmor, eBPF LSM, and controls at the block-I/O layer. A key question is what protection remains effective when the software performing an operation has extensive host privileges.
Storage and Hardware Security
I am interested in data-integrity protections that can provide a stronger trust boundary than host-managed mechanisms alone, including potential hardware-level enforcement.
Selected Project
When AI Agents Go Rogue: Who Protects Your Files?
Keywords: AI Agent Security · Linux Security · eBPF LSM · AppArmor · File Integrity
This project investigates the resilience of sensitive-file protections against a root-privileged AI agent. The experimental design evaluates controls at the file-system, kernel, and block-device layers, and studies the limitations of protections governed by the same host.
Status: Research/poster project. Public artifacts or a citable paper will be linked here when available.
Undergraduate Research
Anhui Polytechnic University
B.Eng. in Electrical Engineering and Automation | September 2021 – June 2025
During my undergraduate studies, I explored the application of intelligent algorithms to engineering problems, with particular interests in computer vision, visual–inertial SLAM and mapping, time-series modeling, and sensor fusion.
- Computer vision and mapping: Studied visual perception, visual–inertial state estimation, sensor fusion, and SLAM-based mapping, as reflected in my coauthored work on monocular visual–inertial SLAM.
- Battery health prediction: Investigated lithium-ion battery health assessment using sample entropy and temporal feature fusion.
- Energy forecasting: Explored time-series modeling and hybrid HP–ARIMA–BP methods for electricity consumption prediction.
- Large models and AI: Gained familiarity with the concepts and applications of large language models (LLMs) and other large-scale AI models, alongside my broader interests in intelligent perception.
I also have an interest in 3D reconstruction and image-based scene understanding. My subsequent coauthored work, Lite3R (2026), reflects this continuing research direction.
Together, these experiences developed my foundation in algorithmic modeling, computer vision, sensor fusion, and intelligent computing systems.
Other Interests
I am also interested in applying machine learning and program analysis to vulnerability detection and security analysis.