An Automated, Customizable Framework for Applying Genetic Algorithm to Generate Test Cases for Web Applications (thesis)

Abstract

Web application testing is an integral part of the web application development process. Faults within a web application can damage a company's reputation and lead to financial losses. Customers will lose confidence if they experience inconvenience. Rigorous testing is necessary to expose faults before production release. Test case generation is a time- and resource-consuming process. Testing requirements increase exponentially with code size, and it might be impossible to exhaustively test any sufficiently complex software. This is specially true of web apps where you have multiple platforms integrating together. In this thesis, I propose the use of genetic algorithm to generate usage-based test cases. Genetic-algorithm-based test case generation requires considerably less resources and is customizable and automated. I modeled usage-based test cases (i.e., user sessions) as components of genetic algorithm, namely genes, chromosomes and genomes, and created a customizable and automated genetic-algorithm-based testing framework. I carried out several sets of experiments, running the genetic algorithm and tuning various parameters to evaluate the effect of each parameter on the resulting generated test suite. Our results show that genetic-algorithm-based test case generation is very cost effective. The test suite is considerably smaller in size compared to the initial collection of user sessions and still maintained high resource coverage.