Applying Sensor Integrity Concepts to Detect Intermittent Bugs in Aviation Software

摘要

This paper assesses the potential benefits of applying online integrity monitoring to aviation software. Traditionally, safety analyses for software components of aviation systems have been conducted differently than safety analyses for electrical and mechanical components. In particular, software verification and validation are typically conducted in advance of system deployment, via exhaustive testing. Though software bugs occasionally impact fielded aviation systems, system fault trees do not typically assign failure probabilities to software components. Moreover, online signal monitoring techniques are not generally applied to detect software-system anomalies, also known as bugs. As software complexity increases, and as interactions between software and physical system components become more complex and situational, it is prudent to consider whether exhaustive pre-service software testing is sufficient to maintain system safety, and whether techniques used to evaluate safety for hardware components (e.g. fault trees and monitoring) might be applied to software components to enhance system safety and mitigate the burden of exhaustive software testing. The primary contribution of this paper is to propose an architecture for online bug monitoring and to quantify its potential to relax requirements for pre-service verification without sacrificing overall system safety.