DNN based Malicious Attack Detection System in Edge Computing

Abstract

Identification of anomalous and malicious traffic in the Internet of Things (IoT) network is critical. This is difficult for Internet of Things security to keep an eye on and prevent unwanted assaults. In the Internet of Things network, machine learning (ML) approach models are offered to block fraudulent communication flows. Several ML models are prone to misclassifying predominantly harmful traffic flows due to insufficient feature selection. The issue must be investigated in order to identify appropriate characteristics for accurate malicious traffic identification in the Internet of Things network. A Deep Neural Network model is suggested as a solution to the problem, which uses the confusion matrix to accurately filter the model using the cross-entropy technique. Different variables have been given to the model to forecast the bot’s trustworthiness in order to address the trustworthiness of IoT devices. In this system, Deep Neural Networks is employed as a deep learning technique. As a result, this model can detect malicious attacks with a high degree of accuracy and consistency.