An LSTM-based novel near-real-time multiclass network intrusion detection system for complex cloud environments

The Internet is connected with everyone for sharing and monitoring digital information. However, securing network resources from malicious activities is critical for several applications. Numerous studies have recently used deep learning-based models in detecting intrusions and received relatively robust recognition outcomes. Nevertheless, most investigations have operated old datasets, so they could not detect the most delinquent attack information. Therefore, the current research proposes the long short-term memory (LSTM)-based near real-time multiclass network intrusion detection system (NIDS) utilizing complex cloud CSE-CICIDSS2018 datasets to secure and detect the network anomalous. The proposed strategy utilizes a random forest algorithm for dimensionality reduction and feature selection. In addition, the selected best suitable features were used in a deep learning-based LSTM model developed for detecting network intrusions. The experimental outcomes reveal that the presented LSTM model obtained 99.66% testing accuracy with 0.12% loss. Thus, the suggested approach can detect network intrusions with the highest precision and lowest rate over the earlier designs.