In this paper we present a novel methodology to help debug memory corruption errors during application debug. In this methodology an optimal balance between hardware and software instrumentation is chosen to check at run-time all memory accesses made by an application. To achieve this balance a set of benchmark applications is first analyzed to determine their memory access patterns. The analysis results are used to make our approach low-cost both from a software performance penalty and a hardware area point-of-view. Experimental results show that our innovative approach typically requires less than 2% of a CPU in silicon area for a less than 1% run-time performance overhead. Our method is both low-cost and applicable to high performance microprocessors as well as time-constrained embedded systems.