Enthusiastic about how the underlying system works and how we can improve the system security with privilege separation.
Memorizer: Lossless Instruction-to-Object Memory Tracing in the Linux Kernel.
Memorizer is a self-contained, low-level tracing framework that represents objects as opaque regions of the address space and dynamically tracks (most) object allocations, data accesses, and function calls. The core insight is a low-level object-centric representation that records detailed lifetime information while linking each operation (call/read/write) with its intended target—Memorizer is the first system to provide lossless instruction-to-object memory tracing in Linux.
ForceShield: Dynamic application guardian for Web, Mobile, and IoT
In IoT devices, the number of processes and network connections seldom change. Attacks usually run as background daemon and listening to network port when needed. We identify the limited process and network ports on an IoT devices and a white-list on the IoT devices. By listening to the process and network events from Linux kernel, we can detect malicious processes and network activities on IoT devices.
Based on the log collected from the Nginx modules, we identify different attacks targeting to the device console for IoT devices. Furthermore, we present what kind of attack, e.g. DoS, code injection, and etc. the devices is suffering from.
To manage the Nginx module in IoT devices, we launch a customized shell that can call our services and limited the number of command avaiable to the external world and limit the attack interfaces.
To prevent the BOT, many of the systems use CAPTCHA to identify a real person. Although the text or number in the image have been distorted, we can still use machine learning or deep learning to recognize the text and number. In this project, I leverage the background of computer vision for image pre-processing and using the SVM to recognize the distorted text and numbers. By doing so, we could let BOT automatically book tickets for us.
In this project, we create a painter that support multiple users to draw on the website synchronously. We use the Firebase as our backend databse for syncing up the moves from the front end users.
PainterHave you ever heard of a cover version and wonder who is the singer? We used signal processing and maching learning technique to identify the singer. This way, you'll never have trouble figuring out your favorite singers!
Bachelor of Computer Science in Information Engineer, 2010 - 2014
Master of Computer Science, Concentratin on Information Security, 2015 - 2017
Candidacy of Ph.D. at Computer Science, Focusing on System Security, 2019 - Now
