An Efficient and Privacy-Preserving Federated Learning Scheme for Flying Ad Hoc Networks

In a flying ad hoc network (FANET), unmanned areal vehicles (UAV) communicate to route the message from the source node to the destination. Often, these UAVs carry sensitive information that needs to be transmitted with full confidentiality without revealing the identity of the sender node. Federated learning (FL) is an emerging machine learning approach that can protect the privacy of nodes in the network. However, implementing FL in FANET is challenging due to the highly dynamic network and limited resources. In the existing distributed FL, clients need to communicate with other clients participating in FL to exchange key information, which increases the overhead significantly. Therefore, we propose a group signature-based federated algorithm that can enhance the protection of the nodes’ identity and significantly reduce the overheads by eliminating the need to exchange key information among the drones. Our simulation results show that the proposed algorithm achieves significantly lower computation cost, communication cost, and signaling overhead than existing works. We also verify the security of our algorithm for various known attacks using AVISPA.