Improved dingo optimisation-based feature selection with optimal deep learning enabled intrusion detection technique on cloud environment

Intrusion detection systems (IDSs) in the cloud must manage significant amounts of data, adapt to dynamic environments, and identify unknown and known attack patterns. Cloud-based IDS use behaviour analysis, signature-based detection, machine learning (ML), and anomaly detection methods. Feature selection (FS) methods help find the key patterns or indicators for detecting malicious activities or anomalies in the cloud infrastructure. This study concentrates on designing and developing improved dingo optimisation-based feature selection with optimal deep learning-enabled intrusion detection (IDOFS-ODLID) technique in the cloud environment. The IDOFS-ODLID technique uses FS and hyperparameter tuning strategies to enhance the intrusion detection rate in the cloud environment. In the IDOFS-ODLID technique, IDOFS technique is mainly designed to select features and thereby improves classification performance. For intrusion, the IDOFS-ODLID technique uses an attention-based bidirectional gated recurrent unit (ABiGRU) approach. At last, the IDOFS-ODLID technique uses a bird swarm algorithm (BSA) for the hyperparameter tuning process. The proposed model is simulated using CICIDS2018 dataset and the results portrayed its promising performance with maximum accuracy value of 99.24% over recent approaches.