Software Partitioning for Protecting High-Performance Code

Modern devices of all kinds are increasingly dependent upon software, and that software increasingly constitutes a distinct military assert and battlefield advantage. Hardware anti-tamper solutions offer substantial protection for software assets, but can be difficult to apply due to architectural constraints leading to performance or cost tradeoffs. Software partitioning can resolve these tradeoffs by allowing the use of mixed environments combining protected and unprotected hardware. Tranditional program slicing techniques can handle simple partitioning tasks but are not well-suited to software involving substantial loop structure, which is often the most sensitive aspect of the software. In this work, we will enhance the powerful loop transformation and mapping framework present in Reservoir's R-Stream compiler to handle software partitioning for loop-oriented codes. This will make partition-based software protect viable for a broad class of high-performance software, such as signal processing.