Program Crash Analysis Based on Taint Analysis

Software exception analysis can not only improve software stability before putting into commercial, but also could optimize the priority of patch updates subsequently. We propose a more practical software exception analysis approach based on taint analysis, from the view that whether an exception of the software can be exploited by an attacker. It first identifies the type of exceptions, then do taint analysis on the trace that between the program entry point to exception point, and recording taint information of memory set and registers. It finally gives the result by integrating the above recording and the subsequent instructions analysis. We implement this approach to our exception analysis framework ExpTracer, and do the evaluation with some exploitable/un-exploitable exceptions which shows that our approach is more accurate in identifying exceptions compared with current tools.