Secure Federated Learning: An Evaluation of Homomorphic Encrypted Network Traffic Prediction

With the increasing level of connectivity to the internet, especially wireless system, network traffic monitoring has become an active field of research. Network traffic analysis has many applications, including in resource allocation or management. However, the growing concern regarding privacy makes it difficult for different entities to share network traffic information. Federated learning and homomorphic encryption have been proposed in previous research as a solution to a secure collaborative analysis, but the practicality as well as a thorough evaluation of the approach have to be explored. This article aims to provide a practical study that could be implemented in real life. Aspects like secure multi-party computation are investigated, which allows organization to use different private keys. In addition, data used for the evaluation are generated in totally different environments. These new features are considered since in practice, companies will not use the same private keys and also network traffic data often come from different type of companies. A detailed evaluation of the approach is also presented.