Empowering healthcare: Secure hand gesture authentication in medical IoT with sEMG

Abstract

Enhancing information security via reliable user authentication in wireless body area network (WBAN)-based Internet of Things (IoT) applications has garnered increasing attention. Traditional biometric methods, like fingerprint recognition, carry significant privacy risks because they cannot be cancelled or changed. Once a biometric template is exposed, it cannot be replaced, leading to potential privacy violations. Addressing these challenges, this study proposes a novel Secure EMG Framework, a cancellable biometric modality using surface electromyogram (sEMG) signals encoded by hand gesture passwords for user authentication. sEMG signals are collected from the forearm muscles, specifically the flexor carpi ulnaris (FCU), during hand gestures, forming a unique and secure biometric token. This proposed method enhances security and reliability through a multi-stage process that involves data capture, pre-processing, feature extraction, and machine learning-based computation of matching scores. A cancellable biometric token is generated through the collection of sEMG data during 16 static wrist and hand movements, increasing authentication diversity and security. To ensure signal clarity within the critical frequency range of 5–500 Hz, a Pure Frequency Hamming Filter is used to reduce noise and artifacts in the raw sEMG data. Key time-domain parameters are then extracted to form a 16-length feature vector, enhancing gesture discrimination. To further improve classification accuracy, a Tuned Boost Perfect Classifier is implemented, addressing overfitting and minimizing errors. The matching score computation enables the evaluation of input and registered signal similarity, allowing users to reset compromised biometric tokens. Experimental results validate the method, achieving an accuracy of 99.72%, an F1-score of 96.0%, and an Equal Error Rate (EER) of 0.0037.