On-Air Hand-Drawn Doodles for IoT Devices Authentication During COVID-19

Abstract

In this paper, a new natural human interaction authentication method is proposed for Internet of Things (IoT) devices. In this method, the user draws a doodle on the air for authentication. On-air drawing refers to virtually drawing free hand-drawn doodle passwords through hand gestures on the air without touching anything that is recommended during COVID-19. This study uses the Google Quick Draw Doodle dataset for password doodles. The proposed method is based on a typical video camera, two lightweight convolutional neural networks (CNNs) and a Kalman filter. The first CNN for hand gesture classification was used to overcome dynamic hand gesture challenges on the air. Second CNN for authentication verification. A Kalman filter was used to correct and smooth the path drawn on the air. Two main goals must be achieved to accept the new authentication method: usability and security. The usability evaluation was based on the ISO 9241-11:2018 standard usability model. The results revealed that the accuracy of the proposed authentication method was 95%, efficiency was 94%, and user satisfaction was acceptable. The evaluation of security was based on two threats related to IoT devices: guessing and physical observation. The results show that the password strength of the proposed authentication method is stronger than the traditional 4-digits PIN password. The proposed authentication method is also resistant to physical observation threats.

Keywords

• Convolutional neural networks (CNNs)
• data augmentation
• Kalman filter
• human-computer interaction (HCI)
• hand gesture recognition
• doodle