ACHIEVE: A Common Evaluation Method for Simultaneous Privacy Protection and Collision Detection Performance of CooperativeCollision Avoidance Architectures for Vulnerable Road Users

2019年，德国有862名弱势道路使用者（VRU），包括行人和骑自行车的人，因道路交通事故死亡，11.6万人受伤。为了在未来减少这些数字，协同防撞系统的部署，其中VRU和汽车驾驶员都配备了一个设备，可以警告他们即将发生的碰撞，似乎是一个很有前途的解决方案。为了实现这样的合作保护系统，许多不同的体系结构是可能的。例如，它们的范围从具有设备到设备（D2D）通信和本地处理的对等（P2P）架构到基于混合通信（蜂窝和D2D）的架构，其中处理分布在汽车、VRU和服务器上。这些不同的体系结构不仅影响碰撞检测性能，而且对用户隐私造成不同的威胁。这是由于在整个体系结构中收集、处理和交换有关这些用户的信息。保护用户隐私不仅从法律的角度来说是强制性的，而且我们预计，这也将有助于未来用户对这项技术的接受。虽然已经提出了不同的架构设计并对其进行了单独评估，但是现有技术不允许确定哪个架构提供最佳的冲突检测性能和隐私保护，即，还没有方法来比较不同架构的冲突性能和/或隐私保护。这导致了我们的项目的主要研究问题：我们如何评估和比较可能的架构的合作防撞系统在碰撞检测和隐私保护方面的性能？因此，在ACHIEVE中，我们将开发一种新的方法来评估和比较基于这两个关键方面的不同架构。为了实现这一目标，我们将首先通过识别多个架构变体并对这些架构进行彻底的隐私威胁分析来构建此提案的基础。我们将进一步分析引入隐私保护解决方案对碰撞检测性能的影响。接下来，我们将重点讨论关键性能指标的定义和分析，这些指标将反映碰撞检测性能和隐私保护，并将集成到我们的通用方法中。此外，为了提供这种新方法，我们将把它应用到感兴趣的架构，并对其性能进行综合评估。通过利用我们的方法和评估结果，可以比较可能的架构变体，从而根据它们在碰撞检测和隐私保护方面的性能做出明智的决定。

In 2019, 862 Vulnerable Road Users (VRUs), including pedestrians and bicyclists, died and 116,000 were injured due to road traffic accidents in Germany. To reduce these numbers in the future, the deployment of cooperative collision avoidance systems, in which both VRU and car driver are equipped with a device that can warn them about upcoming collisions, appears to be a promising solution. For the realization of such a cooperative protection system, many different architectures are possible. For example, they range from Peer-to-Peer (P2P) architectures with Device-to-Device (D2D) communication and local processing to architectures based on hybrid communication (cellular and D2D), with processing distributed over the car, VRU, and server. These different architectures not only influence the collision detection performance, but also cause different threats to users’ privacy. This results from the collection, processing, and exchange of information about these users across the architecture. Protecting users’ privacy is not only mandatory from a legal perspective, but we expect that it will also contribute to the acceptance of this technology by future users. While different architecture designs have been proposed and evaluated in isolation, the state of the art does, however, not allow to determine which architecture(s) offer(s) both the best collision detection performance and privacy protection, i.e. no method to compare the different architectures with respect to collision performance and/or privacy protection is available yet. This leads to the main research question of our project: How can we evaluate and compare the performance of the possible architectures for a cooperative collision avoidance system in terms of both collision detection and privacy protection? In ACHIEVE, we will therefore develop a new method to assess and compare different architectures based on these two key aspects. To reach this goal, we will first build the basis of this proposal by identifying the multiple architecture variants and conducting a thorough privacy threat analysis of these architectures. We will further analyze the impact of the introduction of privacy-preserving solutions on the collision detection performance. We will next focus on the definition and analysis of key performance indicators that will reflect both the collision detection performances and privacy protection and will be integrated in our common method. In addition, to provide this new method, we will apply it to the architectures of interest and conduct a comprehensive evaluation of their performance. By leveraging our method and the results of our evaluation, it will be possible to compare the possible architecture variants and hence take informed decisions based on their performance in terms of both collision detection and privacy protection.