Machine-learning assisted model-implemented fault injection

Moradi, Mehrdad; Denil, Joachim

Title

Author

Moradi, Mehrdad

Denil, Joachim

Abstract

Validation and verification of modern safety-critical systems demand an increasing amount of time and effort as systems become more complicated. Fault Injection (FI) is a well-known testing method that stresses the system in an unusual way to examine system's behavior. Traditional FI methods are not preferable in modern Cyber-Physical Systems (CPS) as they require too much effort to locate critical faults in the system. To tackle this problem, we propose an approach where the Machine Learning (ML) algorithm aids FI by efficiently injecting faults in the model under test automatically. The ML algorithm uses domain knowledge and simulation models at different abstraction levels to predict catastrophic faults, which fail the model's properties.

Language

English

Source (book)

SEDES 2020: Software Engineering Doctoral Symposium Workshop 2020 : proceedings of the 8th SEDES Software Engineering Doctoral Symposium Workshop co-located with 13th International Conference on the Quality of Information and Communications Technology (QUATIC 2020), Online (initially located at Faro, Portugal), September 8-11, 2020

Source (series)

CEUR workshop proceedings ; 2822

Publication

CEUR-WS , 2021

Volume/pages

p. 11-19

Full text (open access)

Licensed under a CC BY Attribution license

Faculty/Department				Faculty of Applied Engineering Sciences

Research group				Co-Design of Cyber-Physical Systems (Cosys-Lab)

Publication type				P1 Proceeding

Subject				Computer. Automation

Affiliation				Publications with a UAntwerp address

Source file

http://ceur-ws.org/Vol-2822

Identifier

Creation

10.03.2021

Last edited

17.06.2024

To cite this reference

https://hdl.handle.net/10067/1759230151162165141