Abstract
In recent decades, security threats have posed a high risk to an organization, which is associated with the proliferation of IoT devices and increasing organizational assets. This ensures that the organization is unaware of the IoT devices' connection to their own network. In such cases, the security and integrity of the network might pose a serious security threat to the network communications. In this paper, the capsule network, which is an improved version of the convolutional neural network (CNN), is used to monitor the network traffic to identify accurately the trusted devices connected to the home network. The inadequacy of CNN in identifying IoT devices during their communication on the network has made the present research choose Capsule Networks (CapsNet) for device identification. The Capsule network carries out the operation in an iterative manner in order to attain improved classification of IoT devices. The activation function used in the capsule network is a squash function that normalizes the magnitude of the vector rather than the conventional use of scalar elements. The outputs of the activation function help to find trusted IoT devices through different capsules, which are formally trained using various concepts. The capsule network performs the identification of IoT devices and classifies trusted and non-trusted devices based on the labeled network traffic data. The simulation is performed by the computation of collected labeled network data from the IoT associated network.