Taint Analysis of Security Code in the KLEE Symbolic Execution Engine

TitleTaint Analysis of Security Code in the KLEE Symbolic Execution Engine
Publication TypeConference Paper
Year of Publication2012
AuthorsCorin, R, Manzano, F
Conference Name 2012 International Conference on Information and Communications Security (ICICS)
Date Published10/2012
PublisherLNCS
Conference LocationHong Kong
AbstractWe analyse the security of code by extending the KLEE symbolic execution engine with a tainting mechanism that tracks information flows of data. We consider both simple flows from direct assignment operations, and (more subtle) indirect flows inferred from the control flow. Our mechanism prevents overtainting by using a region-based static analysis provided by LLVM, the compiler infrastructure machine on which KLEE runs. We rigorously define taint propagation in a formal LLVM intermediate representation semantics, and show the correctness of our method. We illustrate the mechanism with several examples, showing how we use tainting to prove confidentiality and integrity properties.
Refereed DesignationRefereed
PDF (Full text): 
Work Package: 
WP3