Intrusion Detection and Irregularity Analysis in 5g Networks using Deep Learning

Abstract

The proliferation of Internet of Things, or IoT, equipment and the introduction of 5G networks have resulted in an explosion of data that is both copious and highly interconnected. There is an immediate need for strong Intrusion Detection Systems (IDS) designed for IoT ecosystems since this highly linked environment poses serious security risks. In this research, we look at how 5G-enabled IoT environments might benefit from deep learning architectures for intrusion detection system development. It assesses the efficacy of four state-of-the-art models in particular: CNN-BiGRU, TCN + LSTM, CNN-Bidirectional LSTM with Attention and Hierarchical Recurrent Neuronal Networks (HRNN). To find out how well each model can spot irregularities and possible security breaches, we run a thorough comparison study, paying special attention to important performance measures like loss and accuracy. Training performance is best for the TCN + LSTM architecture (with a loss of only 0.03 out of all the models tested), while CNN-BiLSTM + Attention comes in second with 94.2% accuracy & a loss of only 0.06. These results greatly aid in the creation of smart, IDS frameworks powered by deep learning, which improve the robustness and safety of IoT networks in the age of 5G connectivity. Furthermore, the findings provide important information regarding the practical use of these mathematical models for protecting smart environments in the future.