Intrusion Detection System for IoT/Cloud Networks Using Federated Learning and Lightweight Cryptography

This study presents a secure solution that utilizes lightweight cryptography (LWC) and intrusion detection systems (IDS) to safeguard cloud networks and Internet of Things (IoT) from cyberattacks. Federated Learning (FL) is suggested for identifying zero-attacks to guarantee the security of different local IoT networks connected to global server of cloud. The proposed federated learning utilizing data from all local IoT network devices to create a generalized Intrusion Detection System (IDS). Local IoT networks consist of clients that communicate updates to their parameters with a central server located in the global cloud. This server integrates these changes and deploys an improved identification algorithm that is compatible with devices from diverse IoT networks. Following each iteration of federated learning, the cloud server provides each IoT client with the latest model, which they then use to train their respective datasets. In order to safeguard privacy and enhance the overall efficiency of the model, it is necessary to encrypt and authenticate the parameters transmitted between IoT networks throughout each round. The idea suggests using PHOTON-Beetle-AEAD-ENC128 algorithm with a 128-bit key, to secure the connection between cloud servers and IoT networks. The studies done for IDS in the Internet of Things (IoT) and cloud networks, both with and without Federated Learning, indicate the usefulness of the proposed approach. The results show a 10% improvement in accuracy for three local IDS after five rounds of FL. Using PHOTON-Beetle-AEAD-ENC-128 for channel encryption provides enhanced defense against contemporary forms of attacks.