Automatic dataset generation for automated program repair of bugs and vulnerabilities through SonarQube

Abstract

Software maintenance is an important and expensive stage during software development. Most of these tasks are done manually with static code analyzers, but this might change if new Artificial Intelligence approaches were used. For this purpose, huge amounts of data are extremely necessary to achieve a good performance by using traditional Data Science and Deep Learning techniques. Accordingly, this paper presents a software capable of creating, automatically, customizable coding error datasets in JSON format by using the SonarQube static analyzer. Consequently, coding error datasets could be easily created, encouraging new maintenance approaches (e.g., automated program repair through Deep Learning Models).

Keywords

• Artificial intelligence
• Machine learning
• Natural language processing
• Automated program repair
• SonarQube
• Automatic dataset generation