Publication

Abstract : Obfuscation is a popular software transformation to protect intellectual property and avoid reverse engineering. It relies on introducing additional instructions and changing control-flow without affecting program semantics. This introduces overheads in terms of memory, execution time and energy consumption for resource-constrained embedded devices. In this work, we show that these overheads are dependent on three factors: the transformations and their combinations selected, the tool used to effect these transformations and the program workload. In addition, there exists a need for measuring the security of obfuscated code. In this work, we develop a framework for evaluating software obfuscation tools potentially highlighting costs and benefits associated with obfuscation and analyze energy-performance-security trade offs for embedded devices. Our experiments using two popular obfuscation tools, Obfuscation Low-Level Virtual Machine (OLLVM) and Tigress, show that obfuscation could potentially lead to a 5-fold increase in execution time and energy consumption depending on these factors. In addition, Tigress provides more security and simultaneously incurs significant energy consumption compared to OLLVM. Our cost-benefit analysis with respect to energy, performance and security can be used to determine the optimal choice of security measures for resource-constrained environments.