A cascaded federated deep learning based framework for detecting wormhole attacks in IoT networks

Abstract

The growth of the internet over the years has resulted in massive use and spread of the Internet of Things (IoT) in many areas. From home networks to industrial IoT, from medicine to transportation, IoT is everywhere. This massive usage has come with a price as attackers make use of sophisticated techniques and malware to target these heterogeneous IoT networks. Moreover, the low processing power and minimal resources of the IoT devices make security and privacy a major challenge. This weakness has motivated attackers to target IoT networks in a wide variety of ways by launching different types of attacks. One such kind of attack is the Wormhole attack which belongs to the family of routing attacks and disturbs the normal routing and flow of network packets in the IoT network. Many techniques involving rule-based, trust-based, machine learning-driven and deep learning assisted have been devised to counter wormhole attacks. The proposed research work presents a cascaded wormhole detection technique for IoT networks which is based on the federated deep learning technique and a Dynamic Trust Factor (DTF). The DTF is based on two trust attributes, whereas Convolutional Neural Network (CNN) and Long Short Term Memory (LSTM) deep learning models have been trained using the federated approach, which guarantees data security and privacy at the node level. The proposed technique achieves the accuracy of 96% and is lightweight due to the cascaded and federated learning approach.