Cybersecurity Assessment of IoT Networks Using the Connected Random Artificial Neural Networks

In order to determine how secure an Internet of Things (IoT) network is, this article suggests a way to find all the affected devices with IP addresses at the same time. The Associated RNN (ARNN) is a recurrent structure that utilizes a particular Random Neural Network (RNN) design, which consists of two sub-networks that are mutually associated and complement each other. Two separate neurons of the ARNN support opposing opinions for every one of the n gadgets or IP addresses that exist in the IoT network: compromised or not hacked. Pairs of neurons in a fully-connected 2n ARNN structure may learn from real-world information even when power is off. A single overall attack detector is provided by the ARNN, which learns about the interdependencies among network nodes, observes the incoming traffic at each node, and formulates a recommendation for every gadget or IP address in an IoT network using the feature selection (Artificial Bee Colony Optimization) method. This eliminates the need for a separate attack detector at each network node. We go over the ARNN learning algorithm and weight initialization, and then we test it with real-world attack data and compare its results to those of other testing and learning methods. Using ground truth data offline and an improved average metric collected from entering packet traffic for online incremental learning, results are produced for both scenarios. The ARNN surpasses all prior methods by a wide margin when compared to the top state-of-the-art methods.