Probabilistic Approaches for Detecting Query-based Vulnerabilities using Machine Learning for Cybersecurity

Sensitive and personal information of organizations and individuals is placed in databases. If fallen into the wrong hands, they can easily lead to data breaches and financial or reputational damage, which creates a reason to secure these sensitive data. Query-based vulnerabilities pose an important risk to web applications by showing weaknesses in the query structures that prevent them from accessing sensitive information or bringing down functional operations. Traditional rule-based systems fail to identify such dynamic threats because they operate with static patterns that cannot keep track of a changing attack. In response, the proposed framework combines Gaussian Naive Bayes (GNB) with advanced feature engineering for highdimensional query analysis. Captures subtle patterns in the queries and achieves real-time detection with minimal false positives. The results indicate that the GNB approach outperformed other machine learning algorithms in identifying query vulnerabilities and demonstrating its effectiveness. The Naive Bayes model efficiently detects SQL queries. Compared with traditional methods, an accuracy of 98 % is achieved