Software robustness through search for robust runtime implementations

Abstract

Software program robustness is improved by successfully masking memory safety errors in the software program. For instance, at least some memory safety errors in a software program can be masked by using a runtime memory manager that approximates the semantics of an infinite heap memory manager. In one example, an approximation of an infinite heap memory manager is implemented by configuring a parameterized memory manager with parameter values such as padding to be added to each allocation on the heap and the amount of deferment before executing a call to free memory on the heap. Ideal configurations balance expected robustness with costs such as added memory and processing time. Ideal configurations can be identified through systematic search of a coordinate space of selected parameters. Ideal configurations can also be identified by statistically correlating success/failure data collected from execution of deployed instances of the software program to the configuration of the memory managers used therein.