Year: 2025
Type: article
Abstract: Needing protection became the security of digital infrastructure providers, thus even needing centralization of network intrusion detection systems. NIDS is short for it, as cyber threats have matured. This framework, based on recurrent neural networks. RNN is proposed before this time for classifying network traffic data recorded from the standard UNSW-NB15 data set, a complete de facto benchmark data set concerning network intrusion detection. The first step of this platform will perform pre-processing of the network traffic dataset that employs label encoders and a standardization approach in shaping the data before it is trained by a model. The pre-processing system also contains provisions for slicing complete datasets into training and test combinations, the technicalities needed for intelligent evaluation. The core implementation of this framework involves an RNN model-most suitable for keeping temporal dependencies in network traffic, which are crucial in differentiating normal from anomalous behaviours in sequential data. Evaluation metrics will be based on accuracy, precision, recall, and F-score; these are all measures signifying an efficient proposal that classifies network traffic. Moreover, this scaled model was implemented in the cloud for real-time monitoring. Hence research in the improvement of NIDS and the following journey toward its real-world implementation will be paving avenues toward reliable solutions in handling the continually increasing range of cyber security problems.
Institution: IBM (United States)
Language: English
Subfield: Computer Networks and Communications
Field: Computer Science
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W4411798705