Secure E-Voting Using Multimodal Biometric

Online voting systems provide convenience and accessibility but face serious security challenges such as voter impersonation, multiple voting, and spoofing attacks using photographs or prerecorded videos. Conventional authentication mechanisms like passwords, OTPs, or single-image face recognition are insufficient to ensure voter authenticity. To overcome these limitations, this project presents a machine learning–based secure online voting system that uses facial biometric authentication with liveness detection. The proposed system verifies voters through real-time face recognition combined with action-based liveness detection to confirm the presence of a live individual. During voter registration, multiple facial expressions including neutral face, blinking, smiling, and head movements are captured using a webcam. Facial features are extracted using the face_recognition library, and a unique facial encoding is generated and securely stored. A duplicate face detection mechanism based on Euclidean distance comparison is implemented to prevent multiple registrations by the same voter. During the voting phase, the voter is authenticated using live facial verification, followed by continuous camera-based presence monitoring to ensure that the authenticated voter remains present while casting the vote. The system prevents multiple voting by maintaining secure voter metadata and records voting activity without storing candidate information, thereby preserving vote anonymity. The prototype is implemented using Python, Streamlit, OpenCV, NumPy, and machine learning–based facial encodings, making it lightweight and deployable on standard hardware. The results demonstrate that the proposed system effectively reduces spoofing attempts, prevents duplicate registrations, and ensures one-person-one-vote integrity. This project offers a practical and secure framework for online voting, enhancing trust and reliability in digital election systems.