REACTOR: Combining static analysis, testing and reverse engineering to detect software defects

• Autor(a) principal: Alessandro Oliveira Arantes
• Data de Publicação: 2016
• Tipo de documento: Tese
• Idioma: eng
• Título da fonte: Biblioteca Digital de Teses e Dissertações do INPE
• Texto Completo: http://urlib.net/sid.inpe.br/mtc-m21b/2016/04.20.19.30

It is increasingly common the use of computer systems to replace human labor in critical systems, and since these systems have become more autonomous in decision making, they demand a high degree of quality and robustness. INPE develops embedded systems for scientific satellites and stratospheric balloons; consequently, the process of verification and validation require special care in detecting and preventing defects. In terms of complexity and system${'}$s domain in question, these processes consume specialists manpower for a long period. In this scenario, the application of techniques that can automatically support test process provide a significant gain in specialists productivity and efficiency. For this purpose, this work performs the source code reverse engineering in order to support a combination of two V\&V processes, static source code analysis and software testing, in order to detect a wider range of defects. The proposed method, called REACTOR (Reverse Engineering for stAtic Code analysis and Testing to detect sOftwaRe defects), complements the traditional way that static code analyzers work by using dynamic information obtained by an automated test case generator, which combines three different black box techniques, being also possible to infer a set of estimated expected results similar to a test oracle. However, the combination of such techniques is not trivial, especially in terms of tasks that commonly demand some action that are not easily automated. Furthermore, the static analysis by itself can not reveal several types of defects that can only be detected by combining static analysis and dynamic information. The REACTOR method has been implemented in a software tool, also called REACTOR, which exempts from a large manual labors amount from testers by automating the process and basing only on applications source code. In addition, REACTOR was applied to some case studies including one of the space application domain, and it performed better than three other well known static code analyzers.