Network Intrusion Detection System Using Attention-Enhanced Recurrent Neural Network (AERNN)

With the rapid growth of cloud computing, the Internet of Things (IoT), and advanced communication technologies such as 5G, the way people connect across the world has changed significantly. However, this evolution has also increased the risk of large-scale cyberattacks. The business world increasingly relies on interconnected systems with distributed data infrastructures; therefore, network security has become a critical issue. Traditional intrusion detection methods that rely on fixed, rule-based, or signature-driven approaches often fail to detect novel, evolving, or stealthy attack types. These old systems cannot change, improve, or learn. The Internet of Things(IoT), cloud computing, and advanced communication technologies such as 5G has significantly increased the complexity and risk of modern networks. This paper proposes an Attention-Enhanced Recurrent Neural Network(AERNN) for intelligent network intrusion detection. The temporal learning capability of recurrent neural networks is merged with an attention mechanism that dynamically emphasizes the most relevant traffic features in the proposed model. This architecture ensures better detection accuracy, convergence speed, and interpretability with less computational overhead because of proper data preprocessing and feature selection. The AERNN model is evaluated using the CICIDS2017 and CICIDS2018 benchmark datasets under multi approach attack scenarios brute force, infiltration attacks, ditributed denial of service, and botnet. Proposed approach comparison with conventional deep learning models like GRU, CNN and LSTM in which overall detection accuracy was 97.2% presented. These results demonstrate that AERNN offers an efficient, scalable solution for intrusion detection in dynamic network environments.