| Since 20 th century,computer technology has gone through a fast development,software is applied wildly in many areas as a information system carrier which has been a skeleton industry.Software loophole could cause a huge lose especially in some sensitive area like financial,electricity and so on.Program analysis and software test is the main method to make sure the software's validity,correctness and safety.In the development of software test,researchers has presented a lot of method which is static program analysis,automation fuzzing test and so on.Software test also progress with the development of modern software industry and the highly need of software safety.Symbolic execution technology is proposed at 1970 s,as been limited by the hardware and software level,it is not been used wildly.As the development of computer technology recently,symbolic execution has been pay much attention again.Dynamic symbolic execution(DSE)which based on the traditional symbolic execution and combine the symbolic and accurate value could cover the code completely and could generate test cases automatically.DSE has got a huge progress in both academic and industrial circle.Many tools based on DSE is implemented and has made a great achievement.But,DSE still has some problem that not been solved for now,like path explosion,pointer compute,constraints solver and so on.This paper is focus on the DSE's constraints solver problem and propose a method which based on Linear Programming(LP)technology to optimize the constraint solver process.This method could simplify the constraint conditions' complexity and make the follow-up process more efficient.This paper choose the KLEE as the research object and add a new module named LP_OP which implement the LP optimization algorithm.To verify the effectiveness of the LP_OP module,this paper choose some tools and use the KLEE to analysis these tools and generate test cases.We compare the time consumption of the constraint solve process and the whole process before add the LP_OP module and after add it.The experiment turn out that the LP_OP module could promote the efficiency of the constraint solve process. |
