Digital watermarking for virtual physically unclonable function data concealment and authentication.

Area of Science:

• Cybersecurity
• Internet of Things (IoT)
• Applied Cryptography

Background:

• Split-learning-based Virtual Physically Unclonable Functions (VPUFs) in IoT networks face security challenges like eavesdropping and replay attacks.
• Existing security mechanisms often struggle to balance robustness with computational efficiency in VPUFs.
• Resource-constrained IoT environments necessitate lightweight yet effective security solutions.

Purpose of the Study:

• To propose a novel digital watermarking approach to enhance the security of Split-Learning-based VPUFs.
• To address vulnerabilities to eavesdropping and replay attacks without significant hardware or computational overhead.
• To provide a cost-effective and scalable security solution for IoT networks.

Main Methods:

• Utilizing deep learning to generate a watermark embedded in the VPUF's latent representation.
• Simulating Rayleigh fading via Jake's Model to obtain secret channel information for watermark creation.
• Employing an autoencoder to generate a robust latent watermark and embedding it into the VPUF's latent response.
• Implementing dual-factor authentication via watermark and latent response verification.

Main Results:

• Experimental testing confirmed high fidelity, reliability, and unforgability of the watermarking process.
• The proposed method demonstrated no compromise on the VPUF's core performance metrics.
• Successful dual-factor authentication was achieved through simultaneous verification.
• The approach proved effective in enhancing VPUF security against common attacks.

Conclusions:

• The digital watermarking technique significantly strengthens the security of Split-Learning-based VPUFs in IoT networks.
• The proposed solution is cost-effective, scalable, and suitable for resource-constrained IoT devices.
• This research offers a practical method to mitigate eavesdropping and replay attacks on VPUFs.