【作者】 王佳昊；

【导师】 秦志光；

【作者基本信息】 电子科技大学 ， 计算机应用技术， 2007， 博士

【摘要】 目标跟踪作为无线传感器网络最主要的应用之一,由于其使用便捷、成本低廉、隐蔽性和可靠性高等优点,从一开始就成为这一领域的研究热点。但是由于传感器网络大规模、自组织等特点和节点性能方面的限制,使得现有的目标跟踪技术和网络安全技术难以达到实际应用的要求,限制了这一技术的发展。对于大规模目标跟踪传感器网络,如何在有限的节点能量条件下,尽可能提高跟踪结果信息的安全性和精确性是保障系统正常运行的决定性因素。论文以分布式计算技术和信息安全技术为理论基础,以无线传感器网络中的安全目标跟踪算法为主攻对象,详细分析了无线传感器网络中的安全问题并研究了大规模目标跟踪传感器网络系统。在此基础上开展了无线传感器网络的密钥管理算法及目标跟踪算法的研究,内容主要涉及随机密钥预分配策略及其安全性扩展算法、基于动态跟踪簇的分布式跟踪算法、簇会话密钥管理算法、基于非测距的数据融合算法和分布式粒子滤波算法等。对跟踪结果信息的加密和认证是最基本的安全防护策略,论文提出一种多层随机密钥预分配协议H-RPK,来实现大规模无线传感器网络场景中网络层的分布式密钥分配与管理。协议结合了随机密钥预分配、q合成、多路增强、单向密钥序列算法等技术的优点,综合考虑了网络规模、安全性和能量消耗等因素,使其对俘获攻击具有良好的免疫力,可应用于各种大规模无线传感器网络系统。以大规模目标跟踪传感器网络为应用背景,论文提出了一种基于分层结构的目标动态跟踪簇协议DTC,在应用层实现消息驱动的分布式跟踪簇管理和数据融合。通过组织目标周围的传感器节点建立动态跟踪簇来进行本地协作,可以精确定位目标并降低网络的传输负荷,提高网络可靠性和使用寿命。动态跟踪簇能够在逻辑上跟随着目标移动,实现簇自身的维护和更新。在簇首节点上使用改进的凸规划数据融合算法,可以使网络仅依靠简单的非测距传感器就能精确定位目标。仿真证明该算法能有效地跟踪高速移动目标且具有较高的跟踪效率和精确度。论文研究了跟踪簇会话密钥管理策略,在动态跟踪簇协议DTC基础上提出一种簇会话密钥管理策略TCRP。簇会话密钥通过底层加密链路进行分配和传输,来对簇成员节点的跟踪结果信息进行加密,并随着簇状态的变化而更新。以一定的传输消耗为代价进一步提高链路加密密钥的破解难度,增强网络以及跟踪结果的安全和可靠性。目标跟踪系统最主要的目的是为了得到尽可能精确的定位结果。论文研究了分布式bayes滤波技术并引入了分布式粒子滤波器DPF,在已有的目标位置估计结果基础上进一步修正跟踪误差。通过在簇首节点上使用序列贝叶斯滤波器来对跟踪结果进行加工,同时使用滑动定位窗来限制节点计算量,可以使跟踪结果更接近目标运动的真实情况。更多还原

【Abstract】 As one of the primary application proposals of wireless sensor network, target tracking has always been a hotspot in research with a lot of merits, such as convenience, cheapness, concealment and robustness. But the traditional tracking and security mechanisms can not be applied to Wireless Sensor Network directly due to many demerits caused by large scale and distributed architecture and limited capability of sensor nodes. These drawbacks restrict WSN’s application, calling for the necessity of developing high security and accuracy tracking algorithms for large scale tracking WSN under the limitation of low energy support.Based on the distributed computing and security technology, secure target tracking algorithms are studied in this dissertation. And based on the study of security problems and tracking models of large scale WSN, the dissertation focuses on key distribution algorithms and dynamic cluster tracking algorithms. The research includes random predistribution key, distributed target tracking protocol, tracking cluster season key management protocol, range-free convex data fusion algorithm and distributed particle filter algorithm.Encryption and authentication of the tracking results are basic security mechanisms. A hashed random predistribution key protocol is proposed in that it combines the merits of basic random predistribution key, q composition, multi-path reinforcement and one way hash function and it can be used in net layer key distribution and management for large scale WSN. The analysis of it sheds light on the fact that this algorithm can effectively resist node capture attack. With the scalability, security and energy of WSN taken into account, this protocol can be applied to all kinds of large scale WSN.For target tracking application, a layered dynamic tracking cluster protocol is proposed to achieve distributed information driven cluster management and data fusion in application layer. Through organizing the sensor nodes around the target and forming a tracking cluster to follow the target, the protocol can accurately track fast moving objects, decrease communication loads and improve the network’s reliability and service life at the same time. The cluster head can also fuse the member reports according to an improved convex fusion method. By this way, the WSN can accurately locate the target with cheap range-free sensors. Simulation result proves that this protocol can track fast moving objects with excellent efficiency and accuracy.To protect the link layer encryption key, season key management concepts are introduced to the tracking cluster mechanism and a tracking cluster rekey protocol is proposed. The cluster season key can be arranged and transmitted through secure links and updated with the cluster’s upgrade. It is used to encrypt cluster member reports. This method can increase difficulty to key analysis and improve the security of the tracking result under certain transmission energy cost.To further improve the tracking accuracy, the dissertation introduces the sequence bayes filter method and proposes a distributed particle filter model to reprocess the tracking result in cluster heads. The DPF method can efficiently improve the tracking accuracy. And a Localizer window method is proposed to limit the computation cost of nodes. The testing result proves that this algorithm can efficiently decrease the tracking error between the estimated results and the location of target.更多还原