Kalman observers in estimating the states of chaotic neurons for image encryption under MQTT for IoT protocol

Chaotic systems based on artificial neurons present high randomness levels that are desired for applications like data encryption. In this paper, the chaotic systems based on the Hopfield, Cellular, Aihara, and the Rulkov neural models are synchronized and implemented on Raspberry Pi, which allows connectivity to a Machine to Machine (M2M) broker that is available under MQTT for IoT protocol. The process of encryption synchronizes two chaotic systems called publisher and subscriber that are controlled by an M2M broker. The publisher sends data that can be recovered by the subscriber, whose state observers are used to estimating the time series of the chaotic neuron to decrypt the data, increasing at the same time the security of the encrypted message. We show that the classical Kalman filter, its extended version, and the recent novelty, the unscented Kalman filter, all of them are powerful techniques in estimating the states of chaotic neurons. The randomness is evaluated by NIST tests, and the image encryption process is evaluated by performing correlation, histogram, variance, entropy, NPCR, and UACI tests.

Automated summary

This paper discusses the synchronization of chaotic systems based on different neural models and the use of an M2M broker for data encryption. The study shows the importance of techniques like the Kalman filter in estimating the states of chaotic neurons, and evaluates the randomness and image encryption process of chaotic systems.