Options
BDD-based value analysis for X86 executables
Citation Link: https://doi.org/10.15480/882.1507
Publikationstyp
Doctoral Thesis
Date Issued
2017
Sprache
English
Author(s)
Advisor
Referee
Title Granting Institution
Technische Universität Hamburg-Harburg
Place of Title Granting Institution
Institute for Software Systems E-16
Examination Date
2017-12-14
Institut
TORE-DOI
We present an abstract domain for integer value analysis that is especially suited for the analysis of low-level code, where precision is of particular interest.
Traditional value analysis domains trade precision for efficiency by approximating value sets to convex shapes.
Our value sets are based on modified binary decision diagrams (BDDs), which enable size-efficient storage of integer sets.
The associated transfer functions are defined on the structure of the BDDs, making them efficient even for very large sets.
We provide the domain in the form of a library that we use in the implementation of a plug-in for the binary analysis framework Jakstab.
The library and the plug-in are evaluated by comparison to set representations in traditional value analyses and application of the plug-in to CPU2006 benchmarks respectively.
Traditional value analysis domains trade precision for efficiency by approximating value sets to convex shapes.
Our value sets are based on modified binary decision diagrams (BDDs), which enable size-efficient storage of integer sets.
The associated transfer functions are defined on the structure of the BDDs, making them efficient even for very large sets.
We provide the domain in the form of a library that we use in the implementation of a plug-in for the binary analysis framework Jakstab.
The library and the plug-in are evaluated by comparison to set representations in traditional value analyses and application of the plug-in to CPU2006 benchmarks respectively.
Subjects
Ganzzahlanalyse
Binäre Entscheidungsbäume
Integer value analysis
DDC Class
004: Informatik