Using Bayesian Modelling to Predict Software Incidents

TL;DR

This paper explores using Bayesian Belief Networks to predict software defects in safety-critical embedded systems, addressing limitations of traditional fault analysis methods especially in SOTIF environments.

Contribution

It introduces a novel approach applying Bayesian modelling to software defect prediction, focusing on safety-critical and SOTIF contexts.

Findings

Preliminary results show promise in defect prediction accuracy.

Bayesian models outperform traditional fault analysis in SOTIF scenarios.

Early research indicates potential for improving safety assessments.

Abstract

Traditionally, fault- or event-tree analyses or FMEAs have been used to estimate the probability of a safety-critical device creating a dangerous condition. However, these analysis techniques are less effective for systems primarily reliant on software, and are perhaps least effective in Safety of the Intended Functionality (SOTIF) environments, where the failure or dangerous situation occurs even though all components behaved as designed. This paper describes an approach we are considering at BlackBerry QNX: using Bayesian Belief Networks to predict defects in embedded software, and reports on early results from our research.