Transfer Learning Based Bi-GRU for Intrusion Detection System in Cloud Computing

Information security is significantly impacted by intrusion detection systems (IDS), which are considered as a critical security concern in the field of cloud computing (CC). In this study, deep feature guided optimized bidirectional Gated Recurrent Unit (Bi-GRU) neural network based transfer learning (TL) technique is proposed for enhancing cloud security. Initially, the min-max normalization process is performed on input traffic data. Further, the pre-trained-residual neural networks (ResNet) is employed as a deep feature extractor to convert the normalized high dimensional traffic data into low dimensional high sensitive data. Finally, the deep learning (DL) model, Bi-GRU neural network based TL with an artificial hummingbird algorithm (AHO) based bio-inspired algorithm is used to recognize the attack classes. The hyper parameter tuning of Bi-GRU is achieved by AHO based optimization process. The proposed model will be calculated based on some evaluation metrics for UNSW-NB15 dataset, and NSL-KDD dataset. The metrics such as accuracy, recall, false alarm rate (FAR) and precision are the performance measured for proposed method and its efficacy is analysed to describe the superiority. Finally, the attacks found in the cloud are correctly classified with accuracy of 0.992 on NSL-KDD dataset.