A cloud-based hybrid intrusion detection system using feedforward autoencoder and gated recurrent units

Intrusion detection systems (IDS) are used to ensure that cloud infrastructures are not vulnerable to unauthorized access and malicious activity. The research paper proposes a unique intrusion detection algorithm that can be applied to cloud systems based on deep learning techniques. It employs deep learning algorithms, such as Feed Forward Autoencoder to extract features and Gated Recurrent Units (GRU) to classify. GRU model correctly identifies patterns that indicate that one is attempting to compromise the system by examining temporal variation of cloud activity. The Feed Forward Autoencoder is fed raw data; thus, it is more precise in distinguishing between classes. Trained on the KDD Cup 99 Dataset and the CICIDS2017 dataset. It achieves a very high accuracy of 99.86% on the CICIDS2017 dataset and 97.8% on the KDD CUP 99 dataset. The model consists of three GRU layers and dense output. This degree of accuracy demonstrates that the proposed approach can be helpful in identifying various types of intrusions within cloud systems. The intrusion detection system proposed might turn cloud infrastructures safer and more secure by applying deep learning algorithms and large dataset training. The work is part of the current research aimed at creating more accurate and efficient means of detecting sensitive data and resources in the cloud computing environment. This model achieved 99.86% accuracy, 99.85 precision, 99.77 recall and 99.86 F-Score.