Malicious Query Detection Using Lexical Analyzer and Machine Learning

SQL injection attacks continue to pose a serious threat in web security that calls for methods of creative solving. With database systems becoming more dynamic and complex, threat landscapes continuing to change, old techniques often do not work well. To address this issue, the study proposes a framework that involves combining machine learning techniques with a custom lexical analyser to predict normal and abnormal SQL queries that target MySQL data tables. This study uses six machine learning classifiers-Support vector machines, Logistic Regression, Linear Regression, Naïve Bayes, K-Nearest Neighbors and Random Forest for accurately telling apart malicious and non-malicious queries and to identify potential risk of SQL injection, contributing to overall web security.