With the advancement of intrusion attacks, the traditional shallow perceptron model is no longer able to detect the increasingly complex intrusion behaviors. In this paper, big data and deep learning technology are combined and then applied to the field of intrusion detection, firstly, network information is monitored in real time through fuzzy equivalence processing and correlation analysis, and then intrusion detection is carried out on information traffic through deep learning methods. After that, a deep learning model for information security network detection (GAN-LSTM) is constructed on this basis for real-time network security protection. The results show that the average detection rate of the GAN-LSTM model for six types of LDoS attacks is 98.27%. The model can accurately detect the location where the LDoS attack traffic is located when there is a strong normal traffic disturbance between 30 and 90 s. The GAN-LSTM detection system is insensitive to the present number, and its detection rate for five types of attacks, including Slowloris and Slow POST, is >90%. In addition, its detection effect on LDoS attacks that support HTTPS encryption has a mean value of 95.50%. GAN-LSTM can accurately detect the fluctuation amplitude and time domain of the data stream sequence. In addition, the GAN-LSTM model can cluster its each network service into three classes with suitable HTTPS detection when the distance between classes is 0.03-2.94, which can be used for network information security protection in real time.