【作者】 俞波；

【导师】 高传善；

【作者基本信息】 复旦大学 ， 计算机应用技术， 2006， 博士

【摘要】 受到商业应用的驱动，近年来无线自组网出现了几个新的研究分支。本文以无线传感器网络(Wireless Sensor Networks)和车载网络(Vehicular Ad hoc Networks)为研究背景，对这些网络中的若干热点安全问题展开了研究。本文的第一部分工作主要涉及传感器网络的密钥管理和部署方案。由于传感器节点的资源受限，密钥管理是传感器网络中一个十分具有挑战性的问题。本文提出了三个为密钥管理服务的部署模型，并介绍了与之相适应的密钥管理方案。本文的方案中，部署区域被划分为网格，而节点被划分为内部节点和网关节点，并分别部署在网格的中心区域与边界位置。当与适当的密钥管理方案相结合后，本文的部署模型能够有效地提高了网格间节点的连通度以及系统抵抗攻击的能力，并且与传统位置敏感的密钥管理方案相比，减少了部署前节点的配置工作。在第一部分工作中，本文的主要贡献在于尝试了从部署模型的角度帮助增强系统的安全性和部署的灵活性，而传统的密钥管理方案是从密钥分配的角度增强系统的安全性。本文的第二部分工作主要针对传感器网络中的选择传递攻击设计了一种入侵检测方案。在本文方案中，我们提出了多项技术实现有效和可靠的入侵检测，包括：基于检查点的确认机制，随机检查点选择技术，k覆盖确认技术，等。许多传统的传感器网络入侵检测方案一般是基于基站或者基于其他的中央式检测机构。而与许多传统的入侵检测方案不同，本文的检测方案中，每个源节点具有收集来自中间节点报警信息的能力。这种能力使得即使在基站在攻击中“致聋”后，源节点仍然有可能进行检测并做出决策。与多路径传递方案相比，本文方案能有效减小通信开销。在第二部分工作中，本文的主要贡献在于提出了一种检查点确认机制来检测选择传递攻击，在保持较低的通信开销的同时获取较高的检测率。本文的第三部分工作主要针对车载网络中的Sybil攻击提出了使用信号强度分布分析的进行位置验证方法。与传统依赖精确测距的位置方案不同，我们试图通过对节点的信号强度分布进行分析，以确定节点的物理位置，并验证其宣告位置是否和计算位置一致。信号强度测量的方法简单易行，绝大多数的无线电组件如MICA2 Node、802.11、Blue Tooth都提供了信号强度指示RSSI(Received Signal Strength Indicator)功能，使得本文方案能够很低的成本适应车载网络的需求。这种方法充分利用了车载网络所特有的若干特性，例如：车辆移动受到马路拓扑限制，车辆移动的高速性，以及基站的支持，等。我们通过了模拟实验证实了本文方案的可行性和有效性。在第三部分工作中，本文的主要贡献在于考察了使用信号强度分布分析的方法验证车辆物理位置。该方法不适用传统的Ad hoc网络和传感器网络，也未被以往的学者考察过。然而，车载网络所具有的特性，使得在车载网络中使用这种方法成为可能。更多还原

【Abstract】 Driven by commercial interests, several new research branches appear in wireless ad hoc networks. We extend our work on several hot security-related topics in Wireless Sensor Networks and Vehicular Ad hoc Networks.The first part of our work is concerned with key management and deployment strategies in Wireless Sensor Networks. Due to the resource constraints, key management has been a quite challenging problem in Wireless Sensor Networks. We propose three deployment models and introduce the corresponding key management scheme. In our scheme, the deployment field is divided into grids, while sensor nodes are divided into two types, i.e. inner nodes and gateway nodes. When combined with an appropriate key management scheme, our scheme can effectively enhance the connectivity between inter-grid nodes and the security resilience. Our scheme also has the potential to reduce the key-allocation work during the pre-deployment phase, significantly improving the deployment flexibility. Our contribution in this work focuses on the attempt to improve system security and flexibility from the perspective of deployment models, whereas previous schemes are proposed from the perspective of key management.The second part of our work is concerning selectively forwarding attacks in Wireless Sensor Networks. We propose an intrusion detection scheme for this type of attack. In our scheme, we use several techniques to improving the detection performance, including: random checkpoint selection, k-covered acknowledgement, etc. Every source node, in our scheme, has the capability to accumulate alert messages from intermediate nodes. This desirable feature empowers source nodes to have the potential to perform detection and take responses, even when the base station is deafened during attacks. The contribution in this work is that we propose a checkpoint-based acknowledgement scheme for detecting selective forwarding attacks. To our knowledge, little work concerning the same problem has been made before.The third part of our work focuses on Sybil attacks in Vehicular Ad hoc Networks. We propose a detection scheme based on the analysis of signal strength distribution. Unlike traditional position verification techniques, which reply on specific hardware, our scheme tries to estimate the node physical position and then verify whether its claimed position is consistent with its physical position. Our signal strength approachis simple but powerful. Most of existing radio components such as MICA2 nodes, 802.11, Blue Tooth, support RSSI (Received Signal Strength Indicator) very well, making our scheme applicable to Vehicular Networks. The contribution of this work is the attempt to verify node position based on signal strength analysis. This technique would be inappropriate for traditional sensor networks and has not been investigated before. However, the unique properties of Vehicular Network present new opportunities to use this technique to our advantage.更多还原