The rapid increase of Android application creation results in essential malware risks that need suitable detection systems to counteract these threats effectively. A machine learning-based Android Malware Detection System stands as the main component of this project development for classifying applications into benign versus malicious types. An evaluation of malware detection takes place through XGBoost (XGB) and Random Forest (RF) and Decision Tree (DT) algorithms by focusing on network patterns as well as API calls and permissions features. The training together with evaluation process for this system relies on a set of classified Android applications. The model achieves higher performance levels together with fewer selected features which produces simplified operational processes. The selection of the best real-time malware detector depends on F1-score calculations after performing accuracy tests and precision checks with recall validations. This study conducts an orderly examination of basic learning principles and ensemble learning principles to determine their major functions and constraints. Better Android security systems emerge from research results that lead to better malware detection algorithms that both achieve high accuracy and reduce false results. Security experts and developers obtain benefits from research implementation by building safer mobile applications. The research incorporates a machine learning system to improve Android security through specified tactics for detecting new malware threats.