Towards the Isolation of Failure-Inducing Inputs in Cyber-Physical Systems: is Delta Debugging Enough?

Abstract

Cyber-Physical Systems (CPSs) combine digital cyber technologies with parallel physical processes. On the one hand, verification methods of such systems mostly rely on (system level) simulation-based testing. This technique is expensive because complex mathematical models are used to model the physical part of CPSs. On the other hand, test cases of CPSs are usually formed by long test inputs that aim at mimicking real-world scenarios. As a result, when a failure is exhibited, it is highly important to isolate the failure-inducing inputs to provide the developers with the minimal test input. This allows reducing debugging costs by (1) reproducing the failure in the minimal time and (2) reducing the test coverage of the system, making the fault localization easier. In this paper we adapt the well-known delta debugging algorithm to isolate the failure-inducing inputs of CPSs modeled in Simulink. By means of three Simulink models, we analyzed whether Delta Debugging is effective enough to isolate failure-inducing inputs in CPSs.