Fuzzing DBMS via NNLM
Yabin Li, Yuanping Nie, Xiaohui Kuang · 2022
Vulnerabilities in database management system (DBMS) can cause serious security problems affecting hundreds of millions of software systems. Fuzzing is an effective vulnerability mining technology, but few studies utilize neural network language model (NNLM) to fuzz DBMS. In this paper, we explore this technical roadmap and use the sequence model to automatically generate test cases to fuzz DBMS. One of the advantages of the method is that it can effectively test the DBMS in a black-box situation. We implemented our tool NNFuzz and evaluated it on SQLite. Experimental results show that NNFuzz can generate valid test cases and achieve higher code coverage than the initial training set.