Runtime Verification Prediction for Traces with Data

Runtime verification (RV) can be used for checking the execution of a system against a formal specification. First-order temporal logic allows expressing constraints on the order of occurrence of events and the data that they carry. We present an algorithm for predicting possible verdicts, within (some parametric) k events, for online monitoring executions with data against a specification written in past first-order temporal logic. Such early prediction can allow preventive actions to be taken as soon as possible. Predicting verdicts involves checking multiple possibilities for extensions of the monitored execution. The calculations involved in providing the prediction intensify the problem of keeping up with the speed of occurring events, hence rejecting the naive brute-force solution that is based on exhaustively checking all the extensions of a certain length. Our method is based on generating representatives for the possible extension, which guarantee that no potential verdict is missed. In particular, we take advantage of using BDD representation, which allows efficient construction and representation of such classes. The method is implemented as an extension of the RV tool DejaVu.