L-band Digital Aeronautical Communication System (LDACS) is a newly proposed modern state-of-the-art system that will enable communication, navigation, and surveillance in the future aviation network. The current LDACS system does not prevent and detect intrusion within the LDACS domain. Therefore, it may suffer from various cyber-attacks, including spoofing, injection and many more attacks. To the best of our knowledge, this paper proposes the first federated learning-based attack detection model, called FL-Guard, for LDACS. Our proposed model exploits a federated learning environment and uses a deep neural network (DNN) to detect possible attacks on LDACS-based Air-Ground communication. FL-Guardis was simulated on a network of four aeroplanes, and the preliminary results show that the proposed model can detect attacks with 89 % accuracy.