A Comprehensive Review of Deep Learning-Based Intrusion Detection Frameworks for High-Precision Threat Detection in Cloud Infrastructure

Due to the heterogeneous nature of cloud computing environments, it becomes difficult for high-fidelity security monitoring implementations. Classical methods of IDS techniques implementation, which are predominantly signature and rule-based, have all failed to accurately detect emerging and zero-day attacks due to their restricted adaptability and dependence on pre-defined threat signatures. The utilization of automated feature learning and feature capturing of complex but temporal and spatial patterns. These features are a consequence of deep learning technology. The patterns are from raw network flows, system logs, and a host of metrics. In addition, Convolutional Neural Networks (CNNs) can extract structural traffic features, Recurrent Neural Networks (RNNs) and Long Short-Term Memory (LSTM) networks capture sequential dependencies, while transformer-based models leverage self-attention to model global contextual interactions.