The dynamic, multi-tenant nature of cloud computing, coupled with the rapid increase in sophisticated cyberattacks (e.g., distributed denial-of-service, inter-VM attacks), poses unprecedented security challenges [1]. Traditional signature-based Intrusion Detection Systems (IDS) are inadequate for detecting zero-day and anomalous threats inherent in virtualized environments. This paper proposes and evaluates a novel Hybrid Intrusion Detection System (H-IDS) framework utilizing Machine Learning (ML), specifically integrating Convolutional Neural Networks (CNNs) for spatial feature extraction from network flows and Long Short-Term Memory (LSTM) networks for temporal anomaly detection [2]. The proposed H-IDS is deployed as a distributed agent model within a simulated cloud environment (IaaS layer). We detail the feature engineering process, the optimization of the CNN-LSTM architecture, and its performance evaluation against established benchmarks (e.g., Support Vector Machines, shallow Neural Networks) using the large-scale CICIDS2017 dataset and a custom cloud attack scenario [3]. Our results demonstrate that the H-IDS achieves superior performance in terms of detection accuracy, false positive rate (FPR), and robustness against class imbalance, establishing a highly effective mechanism for enhancing proactive cloud security.