| AI Driven Multimodal Authentication Using User Behavior and Multifactor Biometric authentication |
| Paper ID : 1036-ICEEM2025 (R1) |
| Authors |
|
Mohamed Elsayed zaki Elfaramawy *1, Auman Hagahh2, Hisham A. Hamad2, Ashraf Aboshosha3, Ramy N.R Ghaly4 1جامعة حلوان كلية تكنولوجيا التعليم 2Department of Electrical Technology, Faculty of Technology and Education Helwan University 3Rad. Eng. Dept., NCRRT Egyptian Atomic Energy Authority (EAEA) Egypt 4Electric Power Department. Mataria Technical College Egypt |
| Abstract |
| This study evaluates the performance of four machine learning models—Random Forest, Logistic Regression, Support Vector Machines (SVM), and Decision Tree—for anomaly detection in authentication logs. Traditional authentication methods, such as passwords and single-factor biometrics, have proven inadequate in effectively countering the sophisticated and evolving nature of contemporary cyber threats. This inadequacy results in increased vulnerability to attacks such as phishing, brute force attempts, and spoofing, necessitating the development of more robust and adaptive authentication solutions. The hypotheses were evaluated using the following important metrics: AUC-ROC, recall, accuracy, and precision. To fully comprehend model performance, the results highlight the significance of employing a variety of evaluation criteria. The Random Forest model outperformed the others, achieving the highest scores across all metrics, with an AUC-ROC of 94%, highlighting its effectiveness in balancing false positives and false negatives. Additionally, SVM and logistic regression showed steady performance, especially in terms of accuracy and recall, making them strong alternatives in certain scenarios. In contrast, the Decision Tree model exhibited lower performance, suggesting its limited applicability to complex datasets. These findings highlighted the benefit of ensemble approaches in boosting model robustness and offered insightful information on algorithm selection for anomaly detection tasks. To further enhance model performance, future studies could concentrate on fine-tuning hyperparameters, investigating sophisticated ensemble strategies, and evaluating practical implementations |
| Keywords |
| Logistic Regression, AUC-ROC, Support Vector Machines, Real-world implementation |
| Status: Accepted |