Federated learning-based collaborative intrusion detection in highly heterogeneous environments

9thLINCS Scientific Highlights by Gregory Blanc (Télécom SudParis) 

Abstract

Networks of the future, including 5G and Beyond 5G, as well as IoT, are connecting more and more devices to the Internet, improving the connectivity and increasing the service offer, at the expense of their exposure to malicious actors. Although many countermeasures exist, it remains a daunting task to
secure all devices, in particular, when they are owned by third parties. Monitoring has thus gained importance to assist security operators in discovering ongoing threats. Recent works have focused on leveraging machine learning (ML) to automate knowledge acquisition from local data or publicly
available datasets, with the latter focusing on attack data and the former on benign data. The quality of such ML-based intrusion detection system (ML-IDS) depends on the quality or availability of the data. Therefore, a collaborative approach may significantly improve the performance of individual ML-IDSes, provided the data is from similar domains. As local training data cannot be shared for privacy reasons, Federated learning-based IDS (FL-IDS) emerge as a promising solution. A few scientific locks remain on its true ability to enable knowledge sharing, to work on heterogeneous data or to guarantee the privacy of its participants. In this talk, we will discuss some solutions.