HSFI: Accurate Fault Injection Scalable to Large Code Bases

摘要: When software fault injection is used, faults aretypically inserted at the binary or source level. The former isfast but provides poor accuracy while latter cannotscale to large code bases because program must be rebuiltfor each experiment. Alternatives that avoid rebuilding incurlarge run-time overheads by applying decisionsat run-time. HSFI, our new design, injects with all contextinformation from level and applies injectiondecisions efficiently on binary. It places markers in theoriginal can recognized after generation. Weimplemented a tool according design evaluated thetime taken per experiment when using operatingsystems as targets. We perform experiments more quicklythan other source-based approaches, achieving performance thatcome close of binary-level retainingthe benefits source-level injection.