Predicting Integer Overflow through Static Integer Operation Attributes

Integer overflow vulnerability is very difficult to locate and patch. From experience speaking the more complicate the integer operation the more error-prone the program. So in this paper, we come up with a new method to leverage static integer operation attributes to predict integer overflows based on machine learning technique. The static integer operation attributes consist of sink, integer operation accumulation, sanitization and input attributes. Every function of the testing program will be converted to a 10-dimension vector which is fed to several machine learning algorithms to make prediction. Our experiment shows that the method gets a good performance.