Statistical Runtime Checking of Probabilistic Properties

Abstract

Probabilistic correctness is an important aspect of reliable systems. A soft real-time system, for instance, may be designed to tolerate some degree of deadline misses under a threshold.Since probabilistic systems may behave differently from their probabilistic models depending on their current environments, checking the systems at runtime can provide another level of assurance for their probabilistic correctness. This paper presents a statistical runtime verification for probabilistic properties using statistical analysis. However, while this statistical analysis collects a number of execution paths as samples to check probabilistic properties within some certain error bounds, runtime verification can only produce one single sample. This paper provides a technique to produce such a number of samples and applies this methodology to check probabilistic properties in wireless sensor network applications. Abstract. Probabilistic correctness is an important aspect of reliable systems. A soft real-time system, for instance, may be designed to tolerate some degree of deadline misses under a threshold. Since probabilistic systems may behave differently from their probabilistic models depending on their current environments, checking the systems at runtime can provide another level of assurance for their probabilistic correctness. This paper presents a statistical runtime verification for probabilistic properties using statistical analysis. However, while this statistical analysis collects a number of execution paths as samples to check probabilistic properties within some certain error bounds, runtime verification can only produce one single sample. This paper provides a technique to produce such a number of samples and applies this methodology to check probabilistic properties in wireless sensor network applications.

Runtime Verification with State Estimation

“…In this section, we discuss related work on runtime verification of statistical properties and on probabilistic model checking. Sammapun et al [18] consider runtime verification of probabilistic properties of the form: given a condition A, does the probability that an outcome B occurs fall within a given range? Their technique determines statistically, and with an adequate level of confidence, whether a system satisfies a probabilistic property.…”

Frequency Linear-time Temporal Logic

“…Here, however, only a single trace is given from which several finite behaviours of the underlying system are derived to estimate the probability to which extent the formula is satisfied. A notable work in this area is given by [11].…”

Introduction to the special section on runtime verification

“…The system can be partially reset or switched to a failsafe mode. In some situations, for example, when monitoring performance and quality-of-service properties [39], the system can be steered to a more suitable state.…”