Intrusion detection method based on a deep convolutional neural network

This paper presents an intrusion detection method based on a deep convolutional neural network (dCNN) to improve the detection accuracy and efficiency of intrusion detection systems. This method uses deep learning to design the deep intrusion detection model including the tanh, Dropout, and Softmax algorithms. The method first transforms the one-dimensional raw intrusion data into two-dimensional "image" data using data padding. Then, the method uses dCNN to learn effective features from the data and the Softmax classifier to generate the final detection result. The method was implemented on a Tensorflow-GPU and evaluated on a Nvidia GTX 1060 3 GB GPU using the ADFA-LD and NSL-KDD datasets. Tests show that this method has shorter training time, improved detection accuracy, and lower false alarm rates. Thus, this method enhances the real-time processing and detection efficiency of intrusion detection systems. © 2019, Tsinghua University Press. All right reserved.