Runtime Verification with Predictive Semantics

Runtime verification techniques are used to continuously check whether software execution satisfies or violates a given correctness property. In this paper, we extend our previous work of three-valued semantics for Linear Temporal Logic (LTL) to predictive semantics. Combined with the static analysis to the monitored program, the predictive semantics are capable of predicting monitored property's satisfaction/violation even when the observed execution does not convince it. We instrument the monitored program based on its Program Dependence Graph representation in order to emit "predictive word" at runtime. We also implement a prototype tool to support predictive semantics and use it to find predictive words in real, large-scale project. The result demonstrates that the predictive semantics are generally applicable in these projects.