Software Bug Localization with Markov Logic

Software bug localization is the problem of determining buggy statements in a software system. It is a crucial and expensive step in the software debugging process. Interest in it has grown rapidly in recent years, and many approaches have been proposed. How-ever, existing approaches tend to use isolated information to address the problem, and are often ad hoc. In particular, most existing approaches predict the likelihood of a statement being buggy se-quentially and separately. This paper proposes a well-founded, integrated solution to the software bug localization problem based on Markov logic. Markov logic combines first-order logic and probabilistic graphical models by attaching weights to first-order formulas, and views them as templates for features of Markov networks. We show how a number of salient program features can be seamlessly combined in Markov logic, and how the resulting joint inference can be solved. We implemented our approach in a debugging system, called MLNDEBUGGER, and evaluated it on 4 small programs. Our initial results demonstrated that our approach achieved higher accuracy than a previous approach