Optimized Identity Authentication via Channel State Information for Two-Factor User Verification in Information...

Area of Science:

• Computer Science
• Information Security
• Signal Processing

Background:

• Traditional authentication methods like passwords and biometrics are vulnerable to security breaches.
• Existing systems lack robust defenses against forgery, theft, and privacy violations.
• There is a need for advanced authentication techniques to enhance system security and user privacy.

Purpose of the Study:

• To propose a two-factor authentication framework integrating Channel State Information (CSI) with conventional methods.
• To leverage unique CSI variations from user-specific keystroke dynamics for biometric feature extraction.
• To improve the security and reliability of user authentication in information systems.

Main Methods:

• A signal processing pipeline involving Hampel filtering, Butterworth low-pass filtering, and wavelet transform denoising was used.
• Feature extraction included a dual-threshold moving window, subcarrier selection, and Principal Component Analysis (PCA).
• A Kernel Support Vector Machine (SVM) classifier trained with randomized hyperparameter search was employed for classification.

Main Results:

• The proposed method extracts discriminative biometric features from CSI based on keystroke dynamics.
• Signal processing and feature extraction techniques effectively reduced noise and dimensionality.
• The SVM model achieved high accuracy in classifying CSI feature patterns.

Conclusions:

• The integrated CSI and keystroke dynamics framework offers enhanced user authentication security.
• The developed signal processing and machine learning pipeline significantly improves recognition accuracy.
• Experimental results demonstrate a 2-3% improvement in user recognition accuracy compared to current algorithms.