| Abstrakt: |
Network Intrusion Detection System (NIDS) plays a major role in maintaining the integrity and security in computer networks. These systems are created to detect and acknowledge the anomalous activities which specify unauthorized access and malicious internet. Establishing effective NIDS can be difficult, especially identifying network anomalies among the ever-increasing and difficult-to-detect malicious attacks. This study, implemented the Information gain with Deep Bidirectional Long Short-Term Memory (IG-Deep BiLSTM) method utilized to identify the effective intrusions, which will enable an NIDS to gain access to more data. The implemented BiLSTM method can better extract long-term and short-term dependent features and improve classification accuracy. The datasets used to gather data are the ToN-IoT, CIC-IDS-2017, BoT-IoT, and UNSW_NB-15 datasets. Next, pre-processing includes data digitization and encoding, as well as data normalization to convert the actual data into a suitable format and remove noise from the data. The IG is used to select the optimal features, and then the Deep BiLSTM is utilized to classify the network intrusion attacks as normal or malicious. Compared with existing methods, the implemented method achieved high accuracy values of 99.95%, 99.95%, 99.50%, and 99.93% using the CIC-IDS-2017, ToN-IoT, BoT-IoT, and UNSW_NB-15 datasets. [ABSTRACT FROM AUTHOR] |
