Securing IoT: Balancing Privacy and Attack Prediction

Abstract

The huge network of Connected devices that exchange and gather data is known as the Internet of Things (IoT). But this connectedness also creates vulnerabilities, opening up IoT networks to hackers that might steal information, interfere with operations, or even be physically harmful. Our study suggests a novel method that preserve user privacy in IoT network threat prediction by utilizing federated machine learning. Federated learning mitigates privacy problems by enabling models to be trained on dispersed devices without directly sharing sensitive data. The suggested PPIOTN model makes use of the CIC-IOT-2023 dataset, that was created especially for studies on IoT security. Through the use of federated machine learning with differential privacy to train a model on this dataset, the study seeks to secure user privacy while achieving precise cyberattack prediction. Furthermore, proposed PPIOTN architecture’s results are compared with other approaches. Finally, the research is concluded based on tuning the differential privacy parameters and obtaining the satisfied results.