Transforming Network Security by Including Convolutional Neural Networks for Improved Automated Attack Classification and Real-Time Intrusion Detection

Information Science & Engineering

Network intrusion detection systems (NIDS) have been greatly enhanced by developments in machine learning (ML) and deep learning (DL), which have made it possible to analyze network traffic more effectively for anomaly identification. However, the sequential pattern of network transmission is frequently overlooked by current packet-based NIDS, increasing the number of false positives and negatives. Additionally, they usually ignore important header information, which makes it more difficult to identify assaults like denial-of-service (DoS). This research proposes a unique artificial intelligence-enabled paradigm for packet-based NIDS in order to overcome these constraints. By converting sequential packets into two-dimensional images, our technique records temporal linkages in addition to header and payload information. Malicious behavior is successfully identified by the suggested model using convolutional neural networks (CNN). Experiments on publicly accessible datasets show remarkable durability against adversarial instances and high detection rates (97.7%–99%) across a range of attack modes. These outcomes demonstrate our method's potential for precise, real-time intrusion detection in a variety of settings.

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.