Research on Deep Learning of Convolutional Neural Network for Action Recognition of Intelligent Terminals in the Big Data Environment and its Intelligent Software Application

To enhance the accuracy and reliability of big data network traffic intrusion detection, an innovative detection model based on the two-branch convolution and gated recurrent unit was constructed, and its application effectiveness in the field of software engineering was thoroughly explored. This model first performs preprocessing operations on the original network traffic data, and then imports it as input information into the core architecture system jointly composed of the two-branch convolution and gated recurrent unit. Among them, the two-branch convolution focuses on quickly and accurately extracting the multi-dimensional features of the network traffic data, while the gated recurrent unit emphasizes on modeling and analyzing the long-term correlation patterns in the feature sequence. Through the organic combination of the two, the potential features of the data are deeply mined. Subsequently, the obtained feature output is directly transmitted to the classifier, so as to achieve accurate discrimination of various types of network traffic such as normal traffic, DoS attack traffic, DDoS attack traffic, port scanning traffic, and SQL injection attack traffic. The comparison experiment data shows that for network traffic data from different sources and with diverse patterns, the detection accuracy rate of this model is as high as 84.3%. This result strongly verifies its high efficiency and practicality in the field of big data network traffic intrusion detection, provides a solid technical guarantee foundation for the network security protection link in software engineering, has important theoretical and practical values, and is expected to promote the further development and application expansion of network security technologies.