In order to solve the problem that traditional methods cannot achieve a good balance between the accuracy and efficiency of SQL injection behavior detection in the real-time high-speed network traffic environment， this paper proposes a method for real-time detection method of SQL injection behavior based on deep learning construction model, and constructs a detection network model called SQLNN based on Convolutional Neural Networks (CNN) and introduces a fast Fourier transform layer. Based on this model, an online detection and adaptive training framework for SQL injection behavior is proposed. For our detection framework, the detection accuracy of the SQL injection statements reaches 99.98%, and it can detect about 10 000 packets containing SQL statements per second. Therefore, it can satisfy the requirements of real-time online detection of SQL injection attacks for detection accuracy and efficiency.