PPTL specification mining based on LNFG

Abstract

A formal specification is a formal description of system requirements, which is a necessity for formal verification. Due to various reasons, some systems lack of formal specifications. Compared with CTL and LTL, Propositional Projection Temporal Logic (PPTL) has a full regular expressive power. Accordingly, a pattern-library-based PPTL specification mining tool PPTLMiner [1] is able to dig out a better expression of system specification. However, there's still some room for improvement in PPTLMiner, such as deep recursion, high time complexity and space complexity, which limit the use of PPTLMiner in practical applications. Therefore, in this paper, a new PPTL specification mining approach based on Labeled Normal Form Graph (LNFG) is proposed and a supporting tool PPTLMiner+ is also developed. Finally, a specification mining example using PPTLMiner+ is given to illustrate the effectiveness and practicality of the new approach. In this example, a typical algorithm in Swarm Intelligence-Bee Colony Algorithm is selected as a target program, and the working order of EmployedBees, OnlookerBees and ScoutBees is obtained using PPTLMiner+. Further, PPTLMiner+ is compared with PPTLMiner as well as the classical linear specification mining tool Texada [2,3] in terms of time and memory to illustrate the effectiveness and practicability of the proposed method.