This paper presents an innovative approach to detecting and preventing hopping attacks in virtualized environments using Support Vector Machines (SVM) and XGBoost. Hopping attacks occur when attackers exploit vulnerabilities to laterally move between virtual machines (VMs), compromising the security of cloud infrastructures. The proposed solution utilizes SVM for real-time anomaly detection by monitoring VM logs, network traffic, and resource usage, while XGBoost is employed for feature selection, enhancing the detection accuracy. The model identifies unusual inter-VM communication and abnormal resource usage patterns, allowing for immediate countermeasures, such as VM isolation and network restriction, to prevent further lateral movement. The system continuously adapts through a feedback loop, learning from new data and evolving attack patterns. Experimental results demonstrate the effectiveness of this hybrid approach in significantly improving detection rates and minimizing false positives, ensuring robust VM protection. The proposed method offers a scalable and proactive defense against advanced hopping attacks.