On the practical study of IR-based value set analysis

Abstract

There are many potential vulnerabilities in developed programs. Recently, as various programs are developed, the types of vulnerabilities are diversified and the number is also increasing. For this reason, program vulnerabilities analysis techniques are addressed such as dynamic analysis and static analysis. It is difficult to use dynamic analysis because an environment must be set up to execute the target program. In contrast, using static analysis is easy because we only need a program code. However, if we analyze all program flow in static analysis, we need the same cost to analysis as dynamic analysis because we have to execute all instructions like dynamic analysis. For this reason, abstract interpretation is proposed to do static analysis in a limited time. Based on abstract interpretation, value set analysis is studied to infer values that variables can have and to recover them. The primary purpose of this study is to recover variables of a program using value set analysis for vulnerability analysis, address improve memory inference method of value set analysis on IR code, and support all architecture. Additionally, we describe many problems when we developed value set analysis on IR code and how to solve them. We also evaluate the performance of the implemented IR-based value set analysis by comparing it with a state of the art binary analysis tool.