Hardware Embedded Fingerprinting Based on Electric Noise

The world is becoming more globally connected and the number of devices connected to the internet is projected to reach more than 29 billion in the near future and at the core of these devices is for the embedded systems. The authenticity of these embedded system devices will become more crucial to maintain its integrity, security, and the overall infrastructure. Authenticity will ensure that the trusted embedded device is being utilized in the system and will eliminate the counterfeit embedded device that poses security risks for it. To establish the authenticity of embedded devices, we propose in this paper the method of embedded device fingerprinting (EDFP) based on intrinsic characteristics which are the imperfections introduced during the fabrication process, often termed as noise. The main source of noise is the Printed Circuit Board (PCB) as it incorporates all the electronic components. We utilized these intrinsic characteristics to establish an EDFP for Arduino by only using the raw noise extracted from a Pulse Width Modulation (PWM) signal. The extracted noise is then used to train a Deep Neural Network (DNN) model. Experimental results for the proposed solution reveal that from the unique noise associated to each of the Arduinos, it is possible to identify embedded system devices. The overall performance of the trained model for the unseen data is 94%.