Abstract—Structured Query Language (SQL) injection attacks remain one of the most critical web application security threats. Deep Learning (DL) models such as Convolutional Neural Networks (CNN) often require high computational resources, large numbers of parameters, and long training times, which limit their suitability for lightweight detection systems. This study proposes Probabilistic Neural Network (PNN)-based models for SQL injection detection to improve detection efficiency while maintaining low computational complexity. Three PNN-based approaches were investigated: standard PNN, Bayesian Optimization-based PNN (PNN-BO), and Enhanced-PNN with additional dense, batch normalization, and dropout layers. The proposed models were compared with Artificial Neural Network (ANN) and CNN models using a publicly available SQL injection dataset. Experimental results show that the Enhanced-PNN achieved 99.22% accuracy, 99.24% precision, and 99.22% recall while maintaining significantly lower model complexity and training time compared with other DL models. In addition, PNN-BO improved the standard PNN performance from 97.45% to 98.66% accuracy through bandwidth optimization. The findings demonstrate that PNN-based models can provide accurate and computationally efficient solutions for SQL injection attack detection.
 
Keywords—Structured Query Language (SQL) injection, machine learning, deep learning, Probabilistic Neural Network (PNN), Bayesian Optimization (BO)
 
Copyright © 2026 by the authors. This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited (CC BY 4.0).