25 Nov 2021
10:00 Master's Defense Fully distance
Theme
A Fault Injection Framework for Executable UML Models
Student
Guilherme Wilson dos Santos Lopes
Advisor / Teacher
Leonardo Montecchi
Brief summary
Fault Injection (FI) is a widely used verification technique for critical systems. These systems are those that cannot fail, in which defects (or faults) are artificially introduced, to evaluate how the measures implemented to avoid system failures will behave in exceptional conditions. FI can be applied at different levels including physical, hardware and software. Furthermore, FI can also be applied at the model level, although the amount of work in this direction is limited. However, the importance of models in the development of complex systems is growing, making model-level verification and simulations increasingly necessary. The Foundational UML (fUML) is a specific subset of the UML, released as an OMG standard, which has precise execution semantics and therefore can be executed. In this dissertation we applied FI to executable models specified in fUML. First, we characterized the types of faults that can be applied to fUML models, linking them to faults already well established in the literature, and then we implemented an injection mechanism based on automatic model transformation. The methodology was then applied to a simple example in the spatial domain adapted from a real satellite system (NanosatC-BR2). We believe this is an important first step towards early detection of design faults and adoption of IF techniques in fUML models. We conclude by discussing some challenges and directions for future work.
Examination Board
Headlines:
Leonardo Montecchi IC / UNICAMP
Eliane Martins IC / UNICAMP
Maria de Fátima Mattiello Francisco (CSE/ETE)/INPE
Substitutes:
Cecília Mary Fischer Rubira IC / UNICAMP
Leandro Dias da Silva IC / UFAL