A CNN-Based Multimodal Biometric Framework with Continuous Trust Scoring for Secure Web Authentication
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
Traditional web authentication systems face significant challenges from sophisticated spoofing attacks and lack of continuous verification mechanisms.This paper presents a novel CNN-based multimodal biometric framework that integrates face recognition, voice authentication, and liveness detection with continuous trust scoring for secure web authentication aligned with Zero Trust Access Control (ZTAC) principles.The proposed system employs deep Convolutional Neural Net-works (CNNs) for facial feature extraction and anti-spoofing detection, Mel-Frequency Cepstral Coefficients (MFCC) for voice authentication, and real-time liveness detection using temporal analysis.A continuous trust scoring mechanism dynamically evaluates user authenticity throughout sessions, enabling adaptive access control.The framework addresses critical limitations of existing single-modality systems by providing robust defense against presentation attacks including photo spoofing, video replay attacks, and voice synthesis.Experimental validation demonstrates superior performance in detecting spoofing attempts while maintaining usability for legitimate users.The system's web-deployable architecture with low-latency APIs makes it suitable for modern distributed environments requiring stringent security measures.
Fields of Study & Subjects
Biometric Identification and Security
User Authentication and Security Systems
Face recognition and analysis
Publication details
- Journal / source
- Advances in intelligent systems research/Advances in Intelligent Systems Research
- Pages
- 262–275
- Publication date
- 1 April 2026
- Language
- EN
- Document type
- Book chapter