SQL Injection Behavior Mining Based Deep Learning

SQL injection is a common network attack. At present, filtering methods are mainly used to prevent SQL injection, yet risks of incomplete filtering still remains. By deep learning, we detect whether the user behaviors contain SQL injection attacks. The scheme proposed in this article extracts the characteristics of the HTTP traffic in the training sets and uses the deep neural network LSTM and the MLP training data sets, the final predictive capacity of the testing sets is over 99%. The deep neural network uses ReLU as the activation function of the hidden layer, continuously updates the weight parameters through gradient descent algorithm, and finally completes the training within 50 epoch iterations.