Secure eKYC Verification Framework

Abstract

In banking today, traditional KYC processes struggle with security issues, especially being prone to spoofing and robotic attacks, which reduce verification accuracy and make the process less user-friendly. This research presents a solution by integrating blockchain technology with Computer Vision techniques to develop a secure and efficient KYC system. The prototype begins with user registration, requiring personal details such as name, date of birth, Aadhar and PAN numbers, and wallet address. After creating a password, users undergo knowledge-based authentication by answering categorized personal questions. Two options are provided for login: face verification or Aadhar-password login. The face verification method uses a Histogram of Oriented Gradients (HoG) and a Convolutional Neural Network (CNN) to detect and analyze facial features, including eye movement, check if eyes are open or closed, and confirm a blink with an open-closed-open pattern to display the person name, achieving 98% accuracy. After logging in, users provide Aadhar and PAN details for additional verification, followed by liveness detection using a LeNet-5 CNN, confirming real-time eye blinks with 97.5% accuracy. The final step is knowledge-based authentication, where stored answers are verified. Upon successful verification, the system employs the EdDSA algorithm to sign the users identity digitally, ensuring security. The verifier accesses user data via a blockchain oracle, enabling trusted, tamper-proof identity verification.