Research of Intrusion Detection Based on Neural Network Optimized by Sparrow Search Algorithm

In recent years, due to the impact of the COVID-19, most people work from home and study online, resulting in a surge in internet traffic. At the same time, cyber attacks are occurring more frequently. As the second firewall of the system, intrusion detection system can help users discover security threats in time and take corresponding measures through network data monitoring and various alarm mechanisms. To improve the intrusion detection system, a proposal has been made to optimize back propagation neural network using the sparrow search algorithm. This model uses Min-Max scaling and Borderline SMOTE oversampling algorithm to preprocess data, and uses tent map to initialize the population of sparrow search algorithm. Finally, compared with other traditional machine learning models, we choose recall as the core indicator, precision as the secondary indicator, and f1_score as the auxiliary indicator. Experimental results indicate that our model exhibits an improved recall and f1_score, indicating that our model exhibits superior performance in intrusion detection.