An Efficient Deep Learning Framwork for Classification and Detection of Anomaly Based Network Intrusion using NSL-KDD Dataset

The past few years have seen unprecedented growth in the realms of technology and networking, as well as the proliferation of Internet services across all industries and regions. Because of rising instances of piracy and the compromise of several cutting-edge systems, it is crucial that we build information security tools capable of identifying emerging threats. An Intrusion Detection System (IDS) that use deep learning (DL) methods to identify network anomalies is a crucial piece of information security technology. The purpose of this study is to look at DL techniques and see if they can be used in an anomaly-based IDS. The purpose of this research is to employ a hybrid model Gated Recurrent Unit (GRU), Long Short-Term Memory (LSTM), & and (Convolutional Neural Networks (CNN) to identify the unknown assault package in a state-of-the-art IDS with high network performance. Standard classification metrics were used to evaluate DL models trained on the NSL-KDD dataset. These metrics included recall, f1- score, precision, and accuracy of categorization. We used the NSL-KDD dataset to fit and assess the flow. The model's outstanding producing abilities and excellent incursion identification performance are made possible by these data characteristics. Attack detection in the NSL-KDD dataset may be accomplished in two ways: binary classification and multiclass classification. Positive outcomes have been seen when applying the suggested strategy to multiclass classification (98% accuracy) and binary classification (99.99 percent accuracy). IDS have shown encouraging results from experiments using deep IDS models. According on the results of the research, the proposed approach outperforms the competing methodologies.