PHY/MAC layer attack detection system using neuro-fuzzy algorithm for IoT network

Abstract

The Internet of Things (IoT) has become a novel paradigm that includes globally identifiable physical objects, integrated with the internet. This work presents an attack detection model using Artificial Neuro-Fuzzy Interface System (ANFIS) for IoT networks. Based on the input-output profile, ANFIS adapts its rules and membership parameters using hybrid back propagation and learning algorithm. In this paper, Sugeno type ANFIS has been considered. The ANFIS model can take dynamic information such as traffic flow, energy level, packet size, packet rate, source-destination address, source-destination ports, etc. as input profiles, and generate the current network security state as an output profile. The performance of the ANFIS attack detection model can be compared with fuzzy logic, neural networks, and pattern recognition based attack detection models. Performance evaluation shows that the proposed model is more reliable than other approaches based on confusion matrix, mean square error and accuracy measurement.