Qamar Faizan, Kazmi Syed Hussain Ali, Siddiqui Maraj Uddin Ahmed, Hassan Rosilah, Zainol Ariffin Khairul Akram
Center of Cyber Security, Faculty of Information Science and Technology, Universiti Kebangsaan Malaysia, Bangi, Selangor, Malaysia.
James Watt School of Engineering, University of Glasgow, Glasgow, United Kingdom.
PeerJ Comput Sci. 2024 Oct 9;10:e2360. doi: 10.7717/peerj-cs.2360. eCollection 2024.
The emergence of 6G networks promises ultra-high data rates and unprecedented connectivity. However, the effective utilization of the millimeter-wave (mmWave) as a critical enabler of foreseen potential in 6G, poses significant challenges due to its unique propagation characteristics and security concerns. Deep learning (DL)/machine learning (ML) based approaches emerged as potential solutions; however, DL/ML contains centralization and data privacy issues. Therefore, federated learning (FL), an innovative decentralized DL/ML paradigm, offers a promising avenue to tackle these challenges by enabling collaborative model training across distributed devices while preserving data privacy. After a comprehensive exploration of FL enabled 6G networks, this review identifies the specific applications of mmWave communications in the context of FL enabled 6G networks. Thereby, this article discusses particular challenges faced in the adaption of FL enabled mmWave communication in 6G; including bandwidth consumption, power consumption and synchronization requirements. In view of the identified challenges, this study proposed a way forward called Federated Energy-Aware Dynamic Synchronization with Bandwidth-Optimization (FEADSBO). Moreover, this review highlights pertinent open research issues by synthesizing current advancements and research efforts. Through this review, we provide a roadmap to harness the synergies between FL and mmWave, offering insights to reshape the landscape of 6G networks.
6G网络的出现有望带来超高的数据速率和前所未有的连接性。然而，毫米波（mmWave）作为6G预期潜力的关键推动因素，由于其独特的传播特性和安全问题，在有效利用方面面临重大挑战。基于深度学习（DL）/机器学习（ML）的方法成为潜在的解决方案；然而，DL/ML存在集中化和数据隐私问题。因此，联邦学习（FL）作为一种创新的去中心化DL/ML范式，通过在保护数据隐私的同时实现跨分布式设备的协作模型训练，为应对这些挑战提供了一条有前景的途径。在对支持FL的6G网络进行全面探索之后，本综述确定了毫米波通信在支持FL的6G网络背景下的具体应用。因此，本文讨论了在6G中采用支持FL的毫米波通信所面临的特殊挑战；包括带宽消耗、功耗和同步要求。鉴于所确定的挑战，本研究提出了一种名为“联邦能量感知动态同步与带宽优化（FEADSBO）”的前进方向。此外，本综述通过综合当前的进展和研究成果，突出了相关的开放研究问题。通过本综述，我们提供了一个利用FL和毫米波之间协同效应的路线图，为重塑6G网络格局提供见解。
