A Quantitative Safety Assessment Methodology for Safety-Critical Programmable Electronic Systems Using Fault Injection

Given the increased use of programmable embedded electronic systems (PEES) in automotive applications and their vital importance, it is not only important for engineers to design PEES in such a way to meet or exceed safety requirements but also quantify how “safe” these systems are. At the University of Virginia's Center for Safety-Critical Systems, we have developed a safety quantification methodology for embedded real time safety-related systems. The goal of the safety quantification methodology is to provide a generic but rigorous and systematic way of characterizing the dependability behavior of embedded systems that is applicable to a broad range of applications from automotive to nuclear. This paper presents a quantitative safety assessment methodology for safety-critical embedded systems using fault injection (FI). This methodology has been developed, refined and applied to a number of commercial safety-grade systems in the railway, nuclear and avionics industries. Additionally, we present several novel techniques that we developed to overcome long-standing challenges associated with fault injection based safety assessment.