Revolutionary hybrid ensembled deep learning model for accurate and robust side-channel attack detection in cloud computing.

Cryptographic systems are essential for securing sensitive information but are increasingly susceptible to side-channel attacks (SCAs) that exploit physical data leakages. In cloud computing environments, where resources shared across multiple tenants, detecting SCAs becomes particularly challenging due to increased noise and complex data patterns. This study aims to develop a robust detection model for SCAs in cloud environments, leveraging deep learning techniques to capture the multi-dimensional characteristics of power traces while ensuring scalability and accuracy. We propose a hybrid ensembled deep learning (HEDL) model that integrates convolutional neural networks (CNN), long short-term memory (LSTM) networks, and AutoEncoders, enhanced by an attention mechanism to focus on the most critical data segments. The model trained and evaluated on the ASCAD dataset, a benchmark dataset for SCA research, and implemented in a cloud environment to assess real-time detection capabilities. The HEDL model achieved a detection accuracy of 98.65%, significantly outperforming traditional machine learning and standalone deep learning models in both clean and noisy data conditions. The attention mechanism improved the model's focus on key data segments, reducing computational demands and enhancing detection precision. The proposed HEDL model demonstrates superior robustness and accuracy in SCA detection within noisy cloud environments, marking a significant advancement in cloud-based cryptographic security.