A Survey on SQL Injection Detection

The increasing reliance on web-based applications and databases has increased the risk of SQL Injection Attacks (SQLIAs), which is one of the most widespread threats to web platforms. SQLIAs take advantage of weaknesses in user input validation to run malicious SQL queries, leading to unauthorized access, data theft, and extreme economic loss. In response to these issues, many research studies have looked into machine learning (ML) and deep learning (DL) solutions to improve SQLIA detection. Methods like 2D-Convolutional Neural Networks (2D-CNNs), Elastic- Pooling CNNs, Multi-Layer Perceptrons (MLPs), and hybrid approaches like CNN-BiLSTM have shown higher accuracy and adaptability compared to conventional rule-based systems. These models utilize features like word embeddings, syntactic parsing, and vector representations to detect SQLIA patterns efficiently. Using Kaggle and other data sets, experiments show the resilience of DL models in real-time detection and capacity to classify intricate SQLIAs, guaranteeing a substantial improvement in protecting database-driven systems.