Improving the automatic runtime monitor generation process via pruning : a way forward

Abstract

Software quality and robustness are elements that the industry is constantly striving to achieve. However, no guarantees can be given that the end product is bug free. Our research proposes an automated process whereby, existing tests are translated into runtime monitors. This is achieved through a three step process of extracting behavioural knowledge, inferring conditions (modelled as invariants), and finally translating them into runtime monitors. Multiple existing tools such as Daikon have primarily been designed for testing purposes to extract invariants. As a consequence, some invariants inferred have been observed to be unsuitable for the purpose of runtime monitoring; potentially generating false negatives. [excerpt]