Performance analysis of binary code protection

Software protection technology seeks to prevent unauthorized observation or use of applications. Cryptography can be used to provide such protection, but imposes a potentially significant additional computation load. This paper examines the performance impact of two software protection techniques. We develop an analytic model and validate it using a detailed discrete-event simulator applied to memory reference traces of well-known benchmark programs. We find that even though the added workload may be large, that impact is often dominated by inherent costs of disk activity.