Mining Potential Defect Patterns in Unmanned Systems from A Large Amount of Open Source Code

Network software security, as one of the four topics in the road map of unmanned systems in the United States, plays an important role in the technology of unmanned systems, and the defects in the coding level also bring great challenges to the stability of unmanned systems. Many large number of codes vulnerabilities often contained in software arise by using APIs which provided by third-party software, because the developer does not understand the API usage specifications of the third party or cannot accurately understand how to use the API provided by the component in various situations. This paper proposes a defect pattern mining algorithm based on massive open-source codes and implements an automated defect pattern mining tool prototype based on this algorithm. The prototype can analyze the change records of source code, extract the changes involved in API calls, which can find highly relevant method replacement rules and common error fixes, and discover common API coding patterns.