GaitCode: Gait-based continuous authentication using multimodal learning and wearable sensors

The ever-growing threats of security and privacy loss from unauthorized access to mobile devices have led the development various biometric authentication methods for easier safer data access. Gait-based is a popular as it utilizes unique patterns human locomotion requires little cooperation user. Existing gait-based however suffer degraded performance when using such smart phones sensing device, due multiple reasons, increased accelerometer noise, sensor orientation positioning, noise body movements not related gait. To address these drawbacks, some researchers adopted that fuse information sensors mounted on at different locations. In this work we present novel continuous method by applying multimodal learning jointly recorded ground contact force wearable devices. Gait cycles are extracted basic element, can continuously authenticate We use network auto-encoders with early or late fusion feature extraction SVM softmax classification. effectiveness proposed approach has been demonstrated through extensive experiments datasets collected two case studies, one commercial off-the-shelf socks other medical-grade research prototype shoes. evaluation shows achieve very low Equal Error Rate 0.01% 0.16% identification shoes respectively, False Acceptance 0.54%–1.96% leave-one-out authentication.