Tools for Injecting Software Faults at the Binary and Source-Code Level

Abstract

Software Fault Injection (SFI) is an approach for the assessment of fault-tolerant software, which emulates software faults (i.e., bugs) in a software component to assess the impact of these faults on the system behavior and faulttolerance properties. In order to make SFI feasible, support tools are required to reduce the costs of implementing the experimental testbed, to automate and to keep low the duration of SFI experiments, and to oversee the phases of SFI campaigns. The focus of this chapter is on the workflow of SFI campaigns and on their implementation and execution. We detail the steps to be performed in a SFI campaign, in order to allow a tester to reproduce them in SFI experiments, and to highlight the key aspects to which attention should be given when performing SFI. SFI is presented in the context of two SFI tools developed during the CRITICAL-STEP project, namely SAFE, which injects software faults by mutating the source code of the target, and csXception TM /G-SWFIT, which injects software faults by mutating binary code. We also provide a comparison of the two tools in terms of accuracy (i.e., ability to inject faults that match faults in the source code), and lessons learned about the implementation of SFI tools.