Abstract
Ensuring mobile device security is critically important in the context of increasing threats of unauthorized access to confidential data. Traditional authentication methods, such as password protection, PIN codes, and biometric systems, have significant limitations: they can be compromised through social engineering, brute-force attacks, or bypassed if the device is already unlocked. This paper presents an innovative approach to access control based on continuous analysis of contextual characteristics of user behavior. A cross-platform mobile application based on Flutter framework has been developed, integrated with Firebase cloud infrastructure, which collects, processes, and analyzes behavioral patterns of device owners. The system collects parameters such as temporal usage characteristics, geospatial data, network connection parameters, and device interaction profiles. Based on the collected data, a machine learning model is trained that can identify anomalous device usage patterns in real-time. Experimental validation with 25 volunteers over 30 days, collecting 12,500+ sessions, demonstrated that the Random Forest-based system achieves 93.7% accuracy with F1-score of 0.919 and AUC-ROC of 0.965 in detecting unauthorized access while minimizing false positive triggers to 5.8%.