•  
  •  
 

Abstract

Smartphone authentication methods face significant challenges in achieving high accuracy, robustness, and usability within cybersecurity applications. Traditional methods, such as passwords and biometric recognition, often lack adaptability and are prone to high false-positive rates, impacting security and user acceptance. This study presents a novel hybrid approach incorporating machine learning (ML) and the Analytic Hierarchy Process (AHP) in a framework to facilitate decision-making abilities and improve smartphone authentication. A novel dataset was constructed based on 3D touch sensor data (pressure levels and spatial dynamics) collected from 20 participants performing tasks per task over sessions, where AHP was used to rank/choose relevant features. The extracted features were later fed to ML classifiers—such as Random Forest and Support Vector Machine (SVM) components—for user authentication. The hybrid model AHP-ML was extensively evaluated, and it underwent simulated attacks for system resilience testing. As a result, there was a significant difference in the Random Forest model, which achieved an accuracy of 89.7%, precision of 0.88, and recall of 0.90. On the other hand, the SVM model achieved an accuracy of 86.3% with a precision and recall equal to 0.85 and 0.87, respectively. Conclusions AHP-based integration improved classification accuracy by 5–8%, reduced false positives by 4.5% (45 users), and increased legitimate user acceptance of the alarm rate by 6%. The robustness of the model was also validated during attack testing, where it also showed resistance to mimicry and brute-force attacks with a success rate of 3% for mimicry and 1% for brute-force attempts using the Random Forest classifier. The application of AHP in determining feature weighting proves to be a significant step towards achieving an optimal trade-off between security and usability. This AHP augmented machine learning process provides a scalable, flexible solution that strengthens smartphone authentication systems in the context of cybersecurity frameworks and is of great promise for secure and user-friendly mobile application development.

Download

Share

COinS