The Application of Explainable Artificial Intelligence to Low-Power Internet of Things Devices with Secure Communication Using Chaos-Based Cryptography

This paper investigates the feasibility of employing expert knowledge-based Explainable Artificial Intelligence (XAI) for smart house control through low-power Internet of Things (IoT) devices that possess limited computational capabilities. By integrating Explainable AI, we seek to enhance the transparency of the model’s decision-making process, thereby increasing its reliability for the end user. The Arduino Uno board was selected for IoT development because of its extensive popularity and affordability. A model of heating control has been developed using temperature sensors based on the presence of residents in the room. The operational prototype was evaluated by measuring the time taken between data input and decision-making, accompanied by an explanation, to identify any potential bottlenecks that may hinder the performance of the microcontroller. To enhance communication security, we developed a pseudo-random number generation function using chaos-based cryptography with hardware implementation, thus improving communication security without incurring additional computational costs. The method has demonstrated a time efficiency improvement of up to 67% for novice users, 58% for intermediate users, and 50% for expert users.