Location Privacy Protection for UAVs in Package Delivery and IoT Data Collection
Location Privacy Protection for UAVs in Package Delivery and IoT Data Collection
复制标题
DOI：
10.1109/jiot.2023.3293755
复制
发表时间：
2023-12
影响因子：
10.6
通讯作者：
S. Enayati;D. Goeckel;Amir Houmansadr;H. Pishro-Nik
中科院分区：
文献类型：
--
作者：
S. Enayati;D. Goeckel;Amir Houmansadr;H. Pishro-Nik
无人驾驶飞行器（UAV，无人机）因无意或故意侵犯公民隐私而广为人知。然而，从对手观察无人机可推断其目的地这一角度来说，无人机自身也可能成为隐私被侵犯的受害者。本文提出了几种保护隐私的机制（PPM）来保护无人机的位置隐私。特别是，我们针对两种需要截然不同措施的主要无人机应用解决隐私保护问题：1）包裹配送和2）物联网（IoT）数据采集。在包裹配送应用中，我们提出两种不同的PPM来使无人机的飞行轨迹随机化，这样观察的对手就会对无人机的目的地感到困惑；我们提供隐私保障并分析与能耗的权衡。在物联网数据采集场景中，无人机不一定需要精确悬停在物联网设备上方；因此，我们提出一种不同的PPM，根据该机制无人机在物联网设备周围选择一个随机点进行数据采集。然后，考虑最小均方误差（MMSE）准则，我们得出向对手的隐私泄漏情况。我们还分析了网络的平均信息峰值年龄（PAoI），并表明所提出的方法不会显著降低平均PAoI。最后，考虑到MMSE方法在某些应用中的局限性，我们还为这种PPM开发了一种基于差分隐私（DP）的对应方法。我们观察到在拉普拉斯差分隐私中平均PAoI显著下降，但在高斯差分隐私中是可接受的。
Unmanned aerial vehicles (UAVs) are well known for violating citizen’s privacy either inadvertently or deliberately. However, UAVs could be victims of privacy violations themselves in the sense that an adversary observing a UAV can infer its destination. This article proposes several privacy-preserving mechanisms (PPMs) for protecting a UAV’s location privacy. In particular, we address the privacy protection problem in two major UAV applications that require significantly different measures: 1) package delivery and 2) Internet of Things (IoT) data collection. In the package delivery application, we propose two different PPMs to randomize the UAV’s trajectory such that the observing adversary is confused about the UAV’s destination; we provide privacy guarantees and analyze the tradeoff with energy consumption. In the IoT data collection scenario, the UAV is not necessarily required to hover exactly above the IoT device; hence, we propose a different PPM according to which the UAV chooses a random spot around the IoT device for data collection. Then, considering a minimum mean squared error (MMSE) criterion, we obtain the privacy leakage to the adversary. We also analyze the mean Peak Age of Information (PAoI) of the network and show that the proposed method does not degrade the mean PAoI significantly. Finally, considering the limitations of the MMSE approach for some applications, we also develop a differential privacy (DP)-based counterpart for this PPM. We observe that the mean PAoI degrades significantly in Laplacian DP but is acceptable in Gaussian DP.