Comparative Analysis of the Performance of Decision Tree and Random Forest Algorithms in SQL Injection Attack Detection

Abstract

This study compares the performance of two machine learning algorithms the Decision Tree and Random Forest. SQL Injection attacks continue to threaten web applications because they exploit vulnerabilities by injecting malicious code into SQL statements executed on database servers. Therefore, machine learning algorithms are used to identify SQL Injection attacks. The dataset used is 33761 in the form of random query data input in a CSV tabular containing sentence and label columns. The research software used is Google Colaboratory and Microsoft Edge. The series of research conducted by Collect Data is data collection, Preprocessing handling missing values, deleting rows that contain duplicates, and the same query having different labels. Train and Test is used to build models and prepare test data, Build and Compile involves building Decision Tree and Random Forest models. The final step is to evaluate both algorithm models to determine which performs better. After conducting a series of research processes, the results of the Random Forest algorithm are slightly better than the Decision Tree algorithm, with an accuracy of 99.81%, precision of 99.79%, recall of 99.65%, and an average F1-score of 99.72%.

Keywords

• machine learning
• decision tree
• random forest
• sql injection
• database