Detecting Logical Bugs in DBMS via Isomerism Fuzz System
Zhe Wang, Liang Liu, Ning Wang · 2024
The complexity and diversity of DBMSs present significant challenges and implications for testing. Fuzzing has become a standard approach for detecting bugs in DBMSs. However, the increasing complexity of DBMS architectures brings new challenges to its effectiveness and coverage. This paper introduces the idea of fuzzing DBMS via Isomerism System, which includes multiple instances of one DBMS using different components and configurations. By fuzzing these isomorphic systems, the scope of classic differential testing is expanded, increasing its applicability and efficiency. We implemented this system within SQLancer and tested it on popular DBMSs, ultimately discovering 10 previously unknown types of bugs. The experimental results demonstrate the system's usability and efficiency.