DLWIoT: Deep Learning-based Watermarking for Authorized IoT Onboarding

Area of Science:

• Cybersecurity
• Internet of Things (IoT)
• Deep Learning

Background:

• Increasing numbers of IoT devices amplify security risks.
• Current IoT onboarding methods (QR codes, PINs) lack robust protection against unauthorized access and tampering.
• Physical access to devices allows unauthorized onboarding and potential malware installation.

Purpose of the Study:

• To present a novel framework, DLWIoT, for secure and authorized IoT device onboarding.
• To develop a robust, automated image watermarking scheme using deep neural networks for IoT security.
• To enable IoT onboarding exclusively for authorized users by embedding credentials into carrier images.

Main Methods:

• Development of a Deep Learning-based Watermarking for authorized IoT onboarding (DLWIoT) framework.
• Implementation of a deep neural network-based automated image watermarking scheme.
• Embedding user credentials into carrier images, such as QR codes on IoT devices.

Main Results:

• Experimental validation of the DLWIoT framework's feasibility.
• Demonstration of secure IoT onboarding exclusively for authorized users.
• Achieved efficient onboarding times for authorized users within 2.5-3 seconds.

Conclusions:

• DLWIoT provides a robust solution to secure IoT device onboarding against unauthorized access.
• The deep learning-based watermarking approach enhances security without compromising onboarding speed.
• DLWIoT effectively addresses the critical need for authorized access in the expanding IoT ecosystem.