Effective Instrumentation and Runtime Support for Enhancing Software Reliability

The Hong Kong University of Science and Technology
Department of Computer Science and Engineering


PhD Thesis Defence


Title: "Effective Instrumentation and Runtime Support for Enhancing 
Software Reliability"

By

Mr. Hao LING


Abstract:

Modern system and software development have evolved far beyond the 
traditional "code, compile, and run" paradigm. Dynamic analysis and runtime 
monitoring have become essential for enhancing testing, debugging, 
verification, and optimization in continuous development workflows. However, 
standard compilation processes and runtime environments often fail to capture 
the diverse data required for effective dynamic analysis. Integrating 
additional compilation and runtime support addresses this gap, but often 
incurs significant overhead, limiting the scalability and applicability of 
the techniques.

This thesis contributes to improving the scalability and practicality of 
dynamic analysis at an industrial scale. We present three key contributions 
that advance instrumentation~(i.e., the automatic insertion of analysis 
instructions) and the associated runtime support to enhance dynamic analysis 
throughout the entire development lifecycle.

First, GiantSan introduces innovative memory continuity-based instrumentation 
for enhancing the efficiency of memory protections. GiantSan significantly 
improves the accuracy and effectiveness of memory sanitizers, which are 
critical tools for identifying memory-related vulnerabilities. GiantSan 
achieves performance gains by utilizing a novel and efficient interval 
validation algorithm, called Segment Folding, and complementing traditional 
machine instruction-level protections with valuable language- level 
information.

Second, Spinel addresses the challenge of memory-related event monitoring in 
software testing. Existing methods lack efficient mechanisms for extracting 
memory- related guidance in automatic testing. Spinel presents a lightweight 
runtime framework with spatial encoding, capturing essential insights with 
minimal overhead while cooperating with offline analysis to efficiently 
expose hidden bugs.

Third, Zircon provides a compiler-friendly solution to the optimization 
failure issue of event quantification. By refining data flow through 
specialized static analyses and tailored compilation passes, Zircon ensures 
the quality of code optimization with comprehensive instrumentation.

These contributions collectively pave the way for more efficient, accurate, 
and scalable methods tailored for contemporary software systems. Notably, our 
research prototypes have been successfully implemented within a Fortune 500 
company, demonstrating their potential for industrial-scale application.


Date:                   Monday, 18 August 2025

Time:                   9:00am - 11:00am

Venue:                  Room 5504
                        Lifts 25/26

Chairman:               Prof. Nian LIN (PHYS)

Committee Members:      Prof. Charles ZHANG (Supervisor)
                        Dr. Dongdong SHE
                        Dr. Shuai WANG
                        Prof. Jiheng ZHANG (IEDA)
                        Prof. Xiangyu ZHANG (Purdue University)