Abstract
Generally, cloud computing environments (CCE) involve large, dynamic, and high-dimensional data, which makes them an easy target of sophisticated cyberattacks that are difficult to deal with by conventional detection algorithms. To ensure cloud security, this research develops a Quantum-enhanced Variational Stacked Long Short-Term Memory Network (QVSLSTM-Net), a hybrid Deep Learning (DL) framework with quantum processing to achieve better threat detection and minimize false positives on the complex, high-dimensional cloud telemetry. The framework feeds on multi-source cloud logs, which are preprocessed to ensure the data quality, and uses a lightweight temporal encoder to identify significant feature vectors. Now, the significant features are mapped toward a quantum Hilbert space through parameterized quantum feature maps. The Variational Quantum Classifier (VQC) operates on these encodings to make use of quantum-enhanced expressivity to the fine patterns of anomalies, and the stacked LSTM is used as an ensemble of outputs to model the temporal attack signatures. The evaluation shows that the proposed QVSLSTM-Net exposed improvements in AUC (0.98) and minimized detection latency (0.2s) compared to classical baselines such as VQC and LSTM.