Fast and accurate FPGA-based framework for processor architecture vulnerability analysis

• A fast and accurate FPGA-based Processor Vulnerability Analysis platform is proposed.
• The obtained speed-up is about 7 OOM in comparison with simulation-based methods.
• Various scenarios for fault injection and fault propagation are defined and examined.
• AVFs of CPU modules and pipelines with respect to themselves and CPU are provided.
• Using accurate exhaustive scenario, the average of vulnerabilities is less than 15%.

This paper presents a fast, accurate, and flexible FPGA-based fault emulation platform, namely FARAVAM that can be exploited for AVF analysis in modern microprocessors. The proposed approach provides fault injection capabilities supporting automatic modification of post-synthesis net-lists and introduces a highly controllable and observable transient fault analysis environment. The presented vulnerability analysis platform using both exhaustive and random fault emulation approaches, provides useful information for identifying areas threatening reliability to make processors more fault tolerant. We applied our platform for extracting the best trade-offs between precision and speed up in vulnerability analysis of MIPS processor. The experimental results indicate that in addition to having high precision we obtain about seven orders of magnitude speed up in comparison with simulation based vulnerability analysis techniques.