This study investigates the innovative use of keystroke dynamics as a biometric attribute for enhancing smartphone security, leveraging advanced sensors and employing the Column Median approach for template formation. Keystroke patterns were systematically collected from 160 participants across three sessions: one for template formation and two for authentication, with five repetitions in each session. The initial session captured keystroke patterns from various text stimuli, ranging from robust to weak passwords, while subsequent sessions introduced variability with commonplace words and short sentence inputs. Patterns were meticulously collected to emulate real-world usage and rigorously validated through realistic evaluation. A diverse array of anomaly detection methods was employed, identifying the Scaled-Manhattan method as the most effective, with Equal Error Rates (EER) of 2.48% for commonplace words, 2.98% for passwords, and 3.60% for fixed text. This comprehensive approach demonstrates the practical viability of using keystroke dynamics for efficient user authentication, significantly advancing smartphone security and mitigating offline guessing attacks. Future research will refine algorithms, explore larger datasets, and address real-world implementation challenges to enhance the scalability and robustness of this authentication method.