Enhancing Wireless Network Security in 5G using a Deep Learning Approach for Intrusion Detection

Abstract

The recent rapid growth of the 5G and wireless networks has given rise to new security issues as it broadens the attack surface, and intrusion detection becomes more significant. Conventional Intrusion Detection Systems (IDS), such as NIDS are not always effective in the situations involving dynamic traffic and various attack patterns. In this respect, we suggest an IDS framework, which is based on deep learning, integrates unsupervised feature extraction with a Feed-Forward Deep Neural.Network (FFDNN) to classify data and a.Recurrent Neural.Network (RNN) to learn temporal patterns in traffic. The framework is tested on. binary and multiclass classification based on the NSL-KDD data. It proves to be higher in performance than traditional methods of machine learning like SVM, KNN, DT, RF, and NB, as experimental outcomes indicate that it consistently achieves a greater accuracy, precision, recall, and F1 scores. Also, an information gain based feature selection step improves computational efficiency by dimensionality reduction without affecting the detection accuracy. The combination of FFDNN and RNN also allows the system to separate feature-level and sequential dependencies, which are critical in establishing complex attacks. All in all, the results point.out the advantages of deep.learning-based methods in the development of ICS and show the possibility of providing scalable, flexible, and efficient security solutions in 5G wireless and marine communication systems.