Hands-Free Eye Gesture Authentication Using Deep Learning and Computer Vision Principles

Abstract

Authentication systems are security measures that are put in place to safeguard data and systems for users to access a system. Personal identification numbers are used to authenticate and secure users. Password verification using PINs necessitates users entering a physical PIN, which might be subject to password cracking or hacking via shoulder surfing or thermal tracking. To develop an application that will authenticate users more safely and protect the user from shoulder surfing and thermal tracking. The proposed system uses a blend of Deep Learning and Computer Vision techniques to protect the user by combining eye blink count-based PIN entry, and face recognition to avoid shoulder surfing and thermal tracking attacks. Face detection is accomplished by the use of the Haar Cascade classifier, which is a cascade function learned from a large number of positive and negative pictures and is used to recognize objects (faces) in other images, and Dlib’s frontal face detector, which employs the HOG and Linear SVM models. The face is recognized by calculating the features of the face and matching it with the available feature set calculated from the dataset. Eyeblink is counted using the Eye aspect ratio which uses the Euclidean distance formula between the eye coordinates. By taking the count of blinks for every digit, the PIN is generated. PIN authentication with hands-off eye blinks, on the other hand, leaves no physical footprints and thus provides a more secure password entry method when combined with face recognition.