Enhancing security and usability with context aware multi-biometric fusion for continuous user authentication

Abstract

Abstract In this paper, we present a novel continuous authentication system that integrates keystroke dynamics and gait biometrics through a multi-modal fusion framework. The proposed system dynamically adjusts the importance of each biometric modality using the Context-Driven Multi-Biometric Scoring Algorithm (CMBSA), enabling it to adapt to real-time contextual factors such as user behavior and system configuration. Keystroke dynamics are processed using Wavelet Transform Filtering (WTF) to improve feature extraction, while gait data is refined with an Autocorrelation (AC) Filter to ensure the use of reliable gait segments. Experimental results demonstrate that the multi-modal fusion approach significantly enhances authentication accuracy, achieving a combined accuracy of 98.25% and an Equal Error Rate (EER) of 2.35%. The system provides seamless and non-intrusive authentication, ensuring high security and improved usability across different contexts. This research contributes to the development of adaptive, context-aware biometric systems, advancing both security and user experience in real-world applications.