Dynamic Feature Driven Machine Learning Model for Accurate Anomaly Detection in Cloud Environments

Abstract

Cloud environments generate large volumes of dynamic and heterogeneous data, making anomaly detection a challenging task. Traditional static feature-based models often fail to adapt to evolving attack patterns and workload variations. This paper proposes a dynamic feature-driven machine learning model designed to improve anomaly detection accuracy in cloud systems. The proposed model integrates adaptive feature selection with supervised learning algorithms to capture time-varying behavioral patterns. Initially, raw cloud monitoring data is pre-processed and transformed into structured feature sets. A dynamic feature selection mechanism is then applied to identify the most relevant attributes based on statistical significance and temporal variation. The selected features are used to train classification models for distinguishing normal and anomalous activities. Experimental evaluation demonstrates that the proposed approach improves detection accuracy, reduces false alarm rates, and adapts effectively to changing cloud conditions. The results indicate that dynamic feature selection plays a crucial role in enhancing anomaly detection performance in cloud environments.