Dynamic detection approach for integer sign conversion in binary programs
Zhu Xue-me · Computer Engineering and Applications Journal · 2015
For the integer sign conversion faults, which result in buffer overflow vulnerabilities, an approach to detect integer sign conversion faults in binary programs dynamically is presented. Type inference method, based on a framework for binary instrumentation, is used to identify the sign information of integer variables. Then, a set of parameters are obtained,which have a conflict type in memory-related library functions, and are taken as candidate set of potential integer sign conversion faults. The test is run after inserting detection code on the level of intermediate representation to determine the real integer sign conversion faults. The experimental results on Sconvcheck show that this method can effectively detect integer sign conversion faults in binary programs and accurately locate the position the faults have occurred with a lower rate of false positives.