Detecting API-Misuse Based on Pattern Mining via API Usage Graph with Parameters

API misuse is a common issue that can trigger software crashes, bugs, and vulnerabilities. To address this problem, researchers have proposed pattern-based violation detectors that automatically extract patterns from code. However, these detectors have demonstrated low precision in detecting API misuses. In this paper, we propose a novel API misuse detector. Our proposed detector initially extracts API usages from the code and represents them as API Usage Graphs with Parameters (AUGPs). Utilizing the association rule algorithm, it then mines the binary rules, which are subsequently employed to detect the possible violations. The experimental results show that, comparing against five state-of-the-art detectors on the public dataset MuBench, our detector achieves the highest precision (1x more precise than the second-best one) and the highest F1-score (50% higher than the second-best one).