The rapid growth of modern telecommunication networks, driven by 5G & emerging 6G technologies, has enhanced the need for robust, real-time cybersecurity measures to counter threats like DDos, web-attacks, and botnet activities. This paper proposes a lightweight, efficient, anomaly detection system for network edge nodes, utilizing Random-Forest, XGBoost and a hybrid model combining both models using soft voting. The models have been evaluated on the CICIDS2017 dataset, the models achieved a high detection accuracy of 99.85%, 99.88% and 99.86% respectively. The proposed models balance accuracy and computational cost, with low prediction latency and training times, making them suitable for edge deployments. This study demonstrates that ensemble-based ML approaches can deliver high-performance anomaly detection while meeting the demands of real-time telecom network security.