Enhancing Network QoS via Attack Classification Using Convolutional Recurrent Neural Networks

Abstract

Cyber-attacks and intrusions in networks refer to malicious activities that breach or damage data. These activities include direct attacks, such as denial-of-service (DoS) attacks, which overwhelm servers with requests to disrupt services. Intrusion involves unauthorized access to systems by exploiting security vulnerabilities. Malware threats like viruses and worms infect systems to steal information. Additionally, social engineering techniques deceive individuals into revealing sensitive information, while phishing relies on fake messages or websites to gather user data. To prevent these attacks, it is necessary to implement effective security strategies, such as knowing the attack class to protect the network and data. In this paper, ConvRNN (Convolutional Recurrent Neural Network) is used as a large-scale advanced model between Convolutional Neural Networks (CNN) and Recurrent Neural Networks (RNN) to process data containing spatial and temporal information. In addition, ConvRNN generates magical features from data through convolutional layers and serial convolution by RNN, which creates the model's ability to understand complexity, especially in security and surveillance agreements. The simulation results show that the proposed model outperforms LSTM, including precision, recall, F1 score, ROC curve, TPR, FPR, FNR, and accuracy.