In static Keystroke Dynamics (KD)-based user authentication, a fixed-text typing template is periodically updated with patterns for similar text, posing a challenge when users change passwords. Addressing this, we propose a dynamically generated KD template accommodating diverse password types. Leveraging Scaled Manhattan as a classifier, our system achieves a 3.01 ± 1.45% average Equal Error Rate (EER) in practical settings. This study introduces a novel hypothesis, evaluating anomaly detectors’ performance in KD with diverse password texts, ensuring adaptive templates. Utilizing a comprehensive dataset with advanced sensory features from smartphones, covering diverse user profiles, our model surpasses various anomaly detectors. Implications extend to advancing PIN- and password-based authentication, benefiting information system security for modern smartphones.