Federated Learning based Intrusion Detection System for Satellite Communication

With the illimitable urge to colonize an alien planet, extra-terrestrial communication is only possible with the extensive use of planetary satellites in conjunction with terrestrial networks. However, the traditional IP network is overly restrictive as it is dependent on proprietary devices governed by rigid protocols and predominantly they are vertically integrated. Additionally, traditional networks are more prone to various sorts of malicious attacks such as denial of service attacks, which requires security systems to be incorporated while spending an exorbitant sum. Nevertheless, with the use of software defined networking, these limitations can be overcome, as the network orchestration is not constrained by pre-defined hardware with proprietary features or lack of openness and programmability. This openness allows us to integrate various security modules without the need of resorting to expensive security hardware. Therefore, with this experiment we have attempted to implement an extra-terrestrial communication system paired with a federated learning (FL) based intrusion detection system implemented in an inexpensive software defined networking environment. Furthermore, a federated learning based approach ensures enhanced data security over traditional machine learning (ML) based approach and our evaluations show that it is perfectly viable for data centric terrestrial networks.