Trustworthy IoT Infrastructures: Privacy-Preserving Federated Learning with Efficient Secure Aggregation for Cybersecurity

Smart gadgets have become ubiquitous due to the fast development of the Internet of Things (IoT). In an effort to recover the precision and excellence of their services, service providers are aggressively gathering massive amounts of customer data to train machine learning models. This technique, however, has caused people to worry about the security of their personal information. Most importantly, this is accomplished without directly transmitting data to a central server, which greatly reduces the risks of privacy breaches. Users' data stays local in federated learning because machine learning algorithms operate locally, transmitting back only the local classical to the central server. New studies show, however, that local models incorporate user data privacy considerations as well. Additionally, training with existing privacy-preserving secure aggregation approaches requires extremely expensive computing resources or provides inadequate accuracy. Our work presents a secure and efficient aggregation scheme for privacy-preserving federated learning that minimises computational costs. This scheme is well-suited for low-powered Internet of Things devices because it is resilient and fault-tolerant, enabling users to join or leave the system on the fly without negatively impacting the federated learning process or causing abnormal termination. Combining the best features of both methods, this research employed a novel attack classification model based on a hybrid random forest and a Convolutional Neural Network (CNN). Using the CICIoT2023 dataset, we validate the suggested approach and show that it improves the security of IoT ecosystems. Lightweight, trust-managing, and privacy-preserving Internet of Things (IoT) infrastructures are urgently needed, and this research helps to meet that demand.-maintaining answers while cybersecurity threats evolve. Integrating private data into an IoT setting, the suggested system demonstrates trustworthiness.