An Efficient SQL Injection Detection with a Hybrid CNN & Random Forest Approach

Abstract

This study enhances SQL injection attack detection in database-driven applications using a hybrid model combining CNN and RF algorithms. It improves scheduling efficiency, adaptability, and privacy in real-time environments, balancing accuracy and processing time. The CNN model was trained on a comprehensive SQL query dataset for feature extraction, identifying complex patterns. The Random Forest algorithm classified these features to determine if the queries were legitimate or malicious. The model's effectiveness was assessed using correctness, specificity, retrieval, and F-measure. In existing method uses LSTM algorithms for predicting scheduling patterns and optimizing resource allocation based on historical data, incorporating 500 samples to enhance accuracy and reliability. The system employs CNN for feature extraction and RF for classifying and optimizing scheduling decisions. This hybrid method boosts prediction accuracy, scalability, and adaptability. Performance is evaluated on efficiency, processing time, and data privacy, while managing large datasets and multiple agents. The model effectively learns from distributed datasets, reducing processing time from 35 ms to 20 ms. It achieves an accuracy of 95.67% and an error rate of 0.415 Sec, with a significance value of 0.035. The hybrid approach of using CNN and Random Forest for SQL injection detection proves to be efficient and reliable. It offers a strong solution for real-time detection of SQL exploit, consequently Improving web application security measures.