Logo Logo
Hilfe
Kontakt
Switch language to English
Towards cooperative software verification with test generation and formal verification
Towards cooperative software verification with test generation and formal verification
There are two major methods for software verification: testing and formal verification. To increase our confidence in software on a large scale, we require tools that apply these methods automatically and reliably. Testing with manually written tests is widespread, but for automatically generated tests for the C programming language there is no standardized format. This makes the use and comparison of automated test generators expensive. In addition, testing can never provide full confidence in software—it can show the presence of bugs, but not their absence. In contrast, formal verification uses established, standardized formats and can prove the absence of bugs. Unfortunately, even successful formal-verification techniques suffer from different weaknesses. Compositions of multiple techniques try to combine the strengths of complementing techniques, but such combinations are often designed as cohesive, monolithic units. This makes them inflexible and it is costly to replace components. To improve on this state of the art, we work towards an off-the-shelf cooperation between verification tools through standardized exchange formats. First, we work towards standardization of automated test generation for C. We increase the comparability of test generators through a common benchmarking framework and reliable tooling, and provide means to reliably compare the bug-finding capabilities of test generators and formal verifiers. Second, we introduce new concepts for the off-the-shelf cooperation between verifiers (both test generators and formal verifiers). We show the flexibility of these concepts through an array of combinations and through an application to incremental verification. We also show how existing, strongly coupled techniques in software verification can be decomposed into stand-alone components that cooperate through clearly defined interfaces and standardized exchange formats. All our work is backed by rigorous implementation of the proposed concepts and thorough experimental evaluations that demonstrate the benefits of our work. Through these means we are able to improve the comparability of automated verifiers, allow the cooperation between a large array of existing verifiers, increase the effectiveness of software verification, and create new opportunities for further research on cooperative verification.
software verification, formal methods, cooperative verification, software testing, model checking, automated software engineering
Lemberger, Thomas
2022
Englisch
Universitätsbibliothek der Ludwig-Maximilians-Universität München
Lemberger, Thomas (2022): Towards cooperative software verification with test generation and formal verification. Dissertation, LMU München: Fakultät für Mathematik, Informatik und Statistik
[thumbnail of Lemberger_Thomas.pdf]
Vorschau
PDF
Lemberger_Thomas.pdf

6MB

Abstract

There are two major methods for software verification: testing and formal verification. To increase our confidence in software on a large scale, we require tools that apply these methods automatically and reliably. Testing with manually written tests is widespread, but for automatically generated tests for the C programming language there is no standardized format. This makes the use and comparison of automated test generators expensive. In addition, testing can never provide full confidence in software—it can show the presence of bugs, but not their absence. In contrast, formal verification uses established, standardized formats and can prove the absence of bugs. Unfortunately, even successful formal-verification techniques suffer from different weaknesses. Compositions of multiple techniques try to combine the strengths of complementing techniques, but such combinations are often designed as cohesive, monolithic units. This makes them inflexible and it is costly to replace components. To improve on this state of the art, we work towards an off-the-shelf cooperation between verification tools through standardized exchange formats. First, we work towards standardization of automated test generation for C. We increase the comparability of test generators through a common benchmarking framework and reliable tooling, and provide means to reliably compare the bug-finding capabilities of test generators and formal verifiers. Second, we introduce new concepts for the off-the-shelf cooperation between verifiers (both test generators and formal verifiers). We show the flexibility of these concepts through an array of combinations and through an application to incremental verification. We also show how existing, strongly coupled techniques in software verification can be decomposed into stand-alone components that cooperate through clearly defined interfaces and standardized exchange formats. All our work is backed by rigorous implementation of the proposed concepts and thorough experimental evaluations that demonstrate the benefits of our work. Through these means we are able to improve the comparability of automated verifiers, allow the cooperation between a large array of existing verifiers, increase the effectiveness of software verification, and create new opportunities for further research on cooperative verification.