Evaluating Software Diversity in Branch Prediction Analyses for static WCET Estimation

摘要

Static worst-case execution time analysis enables to obtain guaranteed timing bounds for programs, which is required for safety-critical hard real-time systems. This comprises micro-architectural analyses that rely on full knowledge of the executed program. An example are current approaches to statically bound the time penalty of mispredicted branches in systems using static or dynamic branch predictors. On the other hand, in artificial software diversity, uncertainty in program aspects is used to render code-reuse attacks useless, making the system considerably more secure. We solve this conflict by proposing adapted static analyses for static and dynamic branch prediction that are able to cope with diversity, and by quantifying the impact of diversity onto the analysis results through extensive evaluation.