Enhancing Security Against Software Attacks with Reprogrammable Hardware

Security researchers have relied on Field Programmable Gate Arrays (FPGAs) to speed up processes, protect intellectual property, and perform dynamic verification of programs during run-time. The inherent parallelism provided by an FPGA makes these applications extremely efficient, while their re-programmability makes these techniques simple. Until now, FPGAs have been used in security applications solely for these characteristics. Conversely, our research shows that simply running a program on an FPGA can enhance security against software attacks. FPGAs do not rely on memory the same way that general purpose processors (GPPs) do: there is no concept of a stack on an FPGA. By partitioning a vulnerable program between a GPP and FPGA, we show that attacks that rely on the stack can be mitigated by using this partitioning method.