SoFi is an abbreviation of “Survival of the Fittest”, meaning that our approach aims to produce more fitting seeds. SoFi incorporates the reflection mechanism and advanced type inference to generate diverse and semantically meaningful test inputs.
Our approaches leverage semantic feedback by systematically modifying the inputs.
We first proposed the use of dynamic reflection to obtain unseen information and the effect was remarkable.
Type inference, path sensitive analysis, and automatic repair are applied to test cases generation.
These methods lead to more valid test cases, higher coverage, and more bugs found.
This website contains the following pages:
In this page, we describe the developers' fixes for some vulnerabilities.
In this page, we provide related data that reviewers may care about.
In this page, we provide detailed error localization rules .