摘要：

Fault Injection (FI) is a family of techniques that emulates hardware and/or software faults by deliberately inserting them into a system component in order to analyze system behavior under faulty conditions, i.e. whether the system can tolerate faults. It is well recognized that Fault Injection is a powerful means for dependability assessment, especially when the system is composed by third-party components. Due to its importance for supporting dependability evaluation of OTS-based systems in safety-critical domains, it is important to assure the quality of fault injection techniques, i.e accuracy and representativeness properties. This dissertation focused on three fault in- jection techniques employed in systems that integrates OTS components, i.e. Software Fault Injection based on binary code mutation, Interface Error Injection (IEI) and SoftWare-Implemented Fault Injection (SWIFI). Three different problems were addressed: (i) the accuracy of SFI based on code mutations; (ii) the representativeness of existing error models on which IEI techniques are based on; (iii) the representativeness and the effectiveness of existing SWIFI techniques when applied to systems deployed on multicore processors.

展开