Publication Date

8-2023

Date of Final Oral Examination (Defense)

June 2023

Type of Culminating Activity

Thesis

Degree Title

Master of Science in Computer Science

Department

Computer Science

Supervisory Committee Chair

Elena Sherman, Ph.D.

Supervisory Committee Member

Jim Buffenbarger, Ph.D.

Supervisory Committee Member

Steven Cutchin, Ph.D.

Abstract

Accuracy of static analysis over predicate abstract domains depends on the partitions of predicates. More precise predicates approximate concrete values of program variables resulting in more accurate analysis. Manual reasoning about these partitions must be done on a case-by-case basis and is time consuming and difficult.

This work explores learning geometric concepts to automate discovery of predicate domain candidates.

The proposed framework uses run-time data from program executions to gather training data for a PAC-learner to generate separating hyperplanes that can be projected onto predicate domains.

The thesis implements the framework and performs evaluations of it effectiveness on a set of benchmark programs using various test-case generation tools.

This exploratory work discusses several deficiencies in the current state of the art in test case generation, intermediate program representation, and availability of suitable program benchmarks.

DOI

https://doi.org/10.18122/td.2097.boisestate

Share

COinS