The visual information processing technology based on deep learning (DL) can play many important yet assistant roles for unmanned aerial vehicles (UAV) navigation in complex environments. Traditional centralized architectures usually rely on a cloud server to perform model inference tasks, which can lead to long communication latency. Using transfer learning (TL) to unload deep neural networks (DNN) to the edge-fog collaborative networks has become a new paradigm for dealing with the conflicts between computing resources and communication latency. However, ensuring the security of edge-fog collaborative networks entity is still challenging. For such, we propose an anonymous authentication and group key agreement scheme for the UAV-enabled edge-fog collaborative networks, consisting of UAV authentication protocol and collaborative networks authentication protocol. Utilizing the AVISPA assessment tool and security analysis, the security requirements and functional features of the proposed scheme are demonstrated. From the performance results of the proposed scheme, we show that it is superior to existing authentication schemes and promising.

基于深度学习（DL）的视觉信息处理技术可以为复杂环境下的无人机导航发挥许多重要且辅助的作用。传统的中心化架构通常依赖于云服务器来执行模型推理任务，这可能会导致较长的通信延迟。使用迁移学习（TL）将深度神经网络（DNN）卸载到边缘雾协作网络已成为处理计算资源和通信延迟之间冲突的新范式。然而，确保边缘雾协作网络实体的安全仍然具有挑战性。为此，我们提出了一种用于无人机边缘雾协作网络的匿名身份验证和组密钥协商方案，该方案由无人机身份验证协议和协作网络身份验证协议组成。利用AVISPA评估工具和安全分析，论证了该方案的安全要求和功能特征。从所提出的方案的性能结果来看，我们表明它优于现有的认证方案并且有前途。