ECE.11R Dual Biometric – Authentication System
Team Members Heading link
- Irfan Ali
- Cesar Fuentes
- Allan Li
- Nathan Nguyen
Project Description Heading link
Current security systems, including keycard readers and keypads that are widely used on ATMs or company security, had several flaws. Traditional security systems like keycard readers and keypads are vulnerable to exploitation, risking theft of sensitive data. Current options on the market are unreliable due to poor security and the need to memorize passwords. To address this, we developed a security method that gives users security clearance by combining two forms of biometric authentication, fingerprint analysis and gait recognition. Gait is a biometric feature that is defined by how a person walks. The system needs to be secure and reliable such that users can have security clearance easily and free of any malicious attackers. Our initial design used two different inputs: a video camera and fingerprint scanner connected to a Raspberry Pi which are placed next to the device or entrance requiring access. The system uses gait recognition in order to identify the user’s unique method of walking. We used a post-processing Python script to eliminate the background and isolate a person walking and then we used a Convolutional Neural Network (CNN) to train our model to identify a unique feature from a person walking. We then used a matching algorithm to compare feature vectors of two individuals and determined their similarities. Along with gait recognition, our system uses a fingerprint scanner that identifies users by their unique fingerprint signatures. We utilized the Scale-Invariant Feature Transform (SIFT) algorithm to pinpoint distinct, scale and rotation-invariant features of the fingerprint known as keypoints. The fingerprint algorithm had 85% accuracy or higher in determining if the input has a match within the existing database. This ensured consistent results, unaffected by the image’s size or orientation. The combined result of the two forms of verification of the user’s identity, and would allow users access to high-security systems.