Critical-time metric for risk analysis against sharp input anomalies: computation and application case study

TL;DR

This paper introduces a critical-time metric for assessing the safety of controlled systems under sharp input anomalies, providing a computational method and demonstrating its application in risk analysis and defense design.

Contribution

It formulates the critical-time computation as an LMI-based problem for uncertain linear systems and applies it to a case study for validation.

Findings

The critical-time metric effectively assesses system safety margins.

An iterative LMI algorithm estimates critical-time under uncertainties.

Application to a quadruple-tank system illustrates practical utility.

Abstract

This paper investigates the critical-time criteria as a security metric for controlled systems subject to sharp input anomalies (attack, fault), characterized by having high impact in a reduced amount of time (e.g. denial-of-service, attack by upper saturation). The critical-time is the maximal time-horizon for which a system can be considered to be safe after the occurrence of an anomaly. This metric is expected to be useful for risk analysis and treatment (prevention, detection, mitigation). In this work, the computational problem of the critical-time for uncertain linear systems and several classes of sharp input anomalies, depending on the input channel and the set of abnormal signal values, is formulated based on the quadratic constraints (QC) framework, representing sets by the intersection of QC inequalities and equalities. An iterative LMI-based algorithm is then proposed to provide an under-estimate of the critical-time. Finally, the potential of the critical-time as a metric for defense design is illustrated and discussed on the quadruple-tank case study through different relevant scenarios.