Execution Trace Reconstruction Using Diffusion-Based Generative Models

Execution tracing is a critical technique for analysing the behaviour of software systems, enabling several key security and performance analysis tasks. However, missing trace events, often resulting from the resource constraints of tracing tools, can compromise the integrity of trace data and impact subsequent analyses. Solutions for trace reconstruction are notably under explored, and consequently the few existing methods frequently fail to fully utilize contextual information leading to poor performance in complex scenarios. This thesis explores the use of diffusion-based generative models, a class of deep learning techniques that have set new benchmarks in various content generation tasks, for reconstructing incomplete trace event sequences. In the first comprehensive evaluation of diffusion models for this purpose, we test their performance using datasets derived from twelve traces collected across four distinct systems. The models are evaluated under various imputation scenarios, including differing sequence lengths and missing data ratios. Among the models tested, the SSSDS4 model demonstrates superior performance, achieving high accuracy, perfect reconstruction rates, and strong ROUGE-L scores across diverse conditions. These findings underscore the potential of diffusion-based models to accurately reconstruct missing events, thereby maintaining trace integrity and enhancing system monitoring and analysis.

Name thesis:degree_name

M.Sc. Computer Science

Level thesis:degree_level

Masters

Discipline thesis:degree_discipline

Faculty of Mathematics and Science

Department dc:contributor.department

Department of Computer Science

Grantor dc:publisher

Brock University

Year dc:date.issued

2025

Author dc:creator

• Janecek, Madeline

Advisor dc:contributor.advisor

• Ezzati-Jivan, Naser

• TECHNOLOGY::Information technology::Computer science::Computer science

Language dc:language.iso

eng

Handle dc:identifier.uri

https://hdl.handle.net/10464/19301

OAI identifier oai:identifier

oai:brocku.scholaris.ca:10464/19301