Collecting, Analyzing and Archiving Results from Fault Injection Experiments

This paper addresses the issue of the identification of the suitable level of observation (readouts and measurements) to characterize fault injection experiments. In practice, several outcomes can be observed in an experiment, but it is not rare for experimenters to consider only one viewpoint or to rely on the first event observed, in order to diagnose the experiment. In addition, there is not always a single way to assert the faulty behaviors as distinct viewpoints might be considered. Accordingly, an elaborate reflection on the types of readouts and measurements to be collected and recorded is an essential dimension for analyzing the faulty behavior of a target system. Another key aspect concerns the need for archiving the experimental data in a suitable way (featuring sufficient details, still in an exploitable format), so that they can be also useful for extended or alternative analyses. With that in mind, the paper sketches some simple guidelines towards the sharing of experimental results via an open data repository.