Projects

The following is a list of unpublished but substantial projects I worked on during my undergraduate studies. I hope to further develop some of these ideas in the future.

Making Trust Explicit in XOS

The goal of this project was to propose a new operating system or OS subsystem. I designed XOS, a novel operating system that enforces the principle of least privilege (PoLP) at the application level. Specifically, XOS allows processes to access resources based on their trustworthiness, restricting the default access to system resources (e.g., CPU, memory, ports) for each application.

Improving Application Crash Consistency with Symbolic Execution and Fuzzing

A variety of record-and-replay tools examine application crash consistency bugs, but they are limited by the diversity of the workloads used for testing. In this project, we hypothesize that a more diverse and robust workload will allow these tools to find a greater number of application-level crash consistency bugs. To test this hypothesis, we compared the bugs found in three applications found by KLEE-generated workloads against workloads generated by our custom grammar-based fuzzer.

Memery: Analyzing Heap Memory for Fun and Profit

Memory cartography exploits inter-region pointers, which rely on constant offsets within memory regions. However, this technique is less effective in nondeterministic memory layouts, such as those in heaps and stacks. In this project, we developed MEMERY, an algorithm that enhances memory cartography by reconstructing high-level data structures in heap memory. MEMERY assumes system protections like ASLR and no-execute bits, but still effectively detects singly- and doubly-linked structures and loops containing sensitive objects like function pointers and strings. It operates without requiring binary instrumentation or debugging symbols, making it effective in dynamic memory environments.

Reproducing Crash Consistency Experiments with ALICE

The goal of this project was to reproduce one or more experiments from a previously published systems paper. We reproduced experiments from “Application Crash Consistency and Performance with CCFS”, which presents a new file system, CCFS, designed to improve application crash consistency. Using ALICE, we tested the behavior of an application under different filesystem guarantees and compared our findings with those from the original paper.

Splicing Data from a Multi-User Application

The aim of this project was to provide system-level support for applications to efficiently implement “the right to erasure” as outlined in the General Data Protection Regulation. As a starting point, I designed a scalable infrastructure based on a spreadsheet’s computational model to offer developers deletion-aware storage abstractions. The key challenge was designing these primitives to preserve the semantic integrity of data after deletion.