INCREMENTAL FAULT DIAGNOSABILITY AND SECURITY/PRIVACY VERIFICATION

Doctoral thesis, 2020

Dynamical systems can be classified into two groups. One group is continuoustime systems that describe the physical system behavior, and therefore are typically modeled by differential equations. The other group is discrete event systems (DES)s that represent the sequential and logical behavior of a system. DESs are therefore modeled by discrete state/event models.
DESs are widely used for formal verification and enforcement of desired behaviors in embedded systems. Such systems are naturally prone to faults, and the knowledge about each single fault is crucial from safety and economical point of view. Fault diagnosability verification, which is the ability to deduce about the occurrence of all failures, is one of the problems that is investigated in this thesis. Another verification problem that is addressed in this thesis is security/privacy. The two notions currentstate opacity and current-state anonymity that lie within this category, have attracted great attention in recent years, due to the progress of communication networks and mobile devices.
Usually, DESs are modular and consist of interacting subsystems. The interaction is achieved by means of synchronous composition of these components. This synchronization results in large monolithic models of the total DES. Also, the complex computations, related to each specific verification problem, add even more computational complexity, resulting in the well-known state-space explosion problem.
To circumvent the state-space explosion problem, one efficient approach is to exploit the modular structure of systems and apply incremental abstraction. In this thesis, a unified abstraction method that preserves temporal logic properties and possible silent loops is presented. The abstraction method is incrementally applied on the local subsystems, and it is proved that this abstraction preserves the main characteristics of the system that needs to be verified.
The existence of shared unobservable events means that ordinary incremental abstraction does not work for security/privacy verification of modular DESs. To solve this problem, a combined incremental abstraction and observer generation is proposed and analyzed. Evaluations show the great impact of the proposed incremental abstraction on diagnosability and security/privacy verification, as well as verification of generic safety and liveness properties. Thus, this incremental strategy makes formal verification of large complex systems feasible.