Research On Software Behavior Specification Mining Based On Extended Finite State Automaton

Model inference of software behavior is an important work in software development.Model inference algorithm helps software developers,testers and operators analyze and debug software systems in different situations.Finite state automaton(FSA)is often used to represent software behavior model.Unfortunately,event parameters often affect the structure of behavior model.Therefore,in order to express behavior model more accurately,existing methods often extend FSA inference algorithms to Extended finite state automaton(EFSA)inference algorithms.However,the existing EFSA inference algorithms are difficult to meet the problem of different use scenarios.It is often inefficient to extend the existing FSA algorithm to EFSA algorithm,which requires a lot of professional rework design and development.How to efficiently extend the algorithms of FSA becomes an important problem.At the same time,the EFSA model inference algorithms are lack of flexibility.We will analyze and solve the problem of control flow angle flexibility.In this paper,we propose a declarative model inference method called PropertyModel inference tool(Prop Mint):(1)Prop Mint can declare the FSA inference algorithm as an EFSA inference algorithm.Prop Mint is applied to two advanced FSA model inference algorithms k Tails and Synoptic,which are declared as EFSA inference algorithms with event parameters.(2)This paper decouples the inference algorithm of process EFSA model,so that users can freely combine property types and property instances,and solves the problem of inflexibility of the inference algorithm of EFSA from the aspect of control flow for the first time.(3)The mining and merging process of process model inference is optimized by the declarative method,which simplifies the inference process on the premise of meeting the needs of users.It greatly improves the efficiency of software behavior model inference.It provides a theoretical basis for dynamic analysis of software behavior.Through experiments,we verify that the declarative method of Prop Mint is better than the procedural method in most cases.It is also found that the efficiency of declarative Prop Mint algorithm is better than that of procedural model inference algorithm when the event type is less and the event log is larger.In the experiment,we proved that the operating efficiency of the k Tails and Synoptic algorithms declared by Prop Mint is significantly higher than that of the procedural k Tails and Synoptic.At the same time,we analyzed the reasons why the declarative model inference algorithm is efficient.