【作者】 林驰；

【导师】 吴国伟；

【作者基本信息】 大连理工大学 ， 计算机应用技术， 2013， 博士

【摘要】 安全关键无线传感器网络要求无线传感器网络具有严格的可靠性、实时性和安全性保证,确保数据高效、及时、可靠、安全的传输。其潜在应用涵盖工业过程监测控制、智能交通管理系统及智能医疗等安全关键系统,具有重要而广泛的应用前景。安全关键无线传感器网络的动态性、运行环境的开放性等因素为无线传感器网络高效可信协议设计带来了巨大挑战,设计协议时需要考虑资源有限性、网络拓扑随机性等不确定性因素,克服能耗分配不均衡、网络生存时间短、安全评估性不足等缺点,进而设计出高效、实时、安全、可靠的无线传感器网络协议体系架构,本论文着重解决和深入探讨了传输高效性、网络容错性、网络安全性和隐私保护性四方面存在的主要问题。四个方面具体研究内容如下：(1)基于蚁群优化的高效数据收集协议。数据收集协议高效性影响着无线传感器网络的能耗分布,直接关系到网络生存时间的长短。现有方法缺乏动态拓扑结构设计和维护的有效方法,导致维护拓扑结构能耗较大,另外还存在节点能耗不均的现象。本文设计了一类基于蚁群优化的无线传感器网络高效数据收集算法(Data Aggregation based on Ant Colony Algorithm, DAACA),将蚁群优化理论与数据收集路由协议相结合。运用蚁群优化中信息素指导收集节点和下一跳节点的选取,进而建立能量感知的动态网络拓扑结构,均衡节点间能耗,减少传输跳数。设计信息素调整方法,动态调整网络拓扑结构,不断优化数据收集树。在此基础上,设计了三种启发式算法,从全局、信息素阈值和蚁群系统三个角度优化信息素的调整方法,进一步实现高效传输和能耗节约。实验结果表明,DAACA的网络生存时间更长、平均能耗更低、数据包传输跳数更少。(2)跳跃式传输实时容错路由协议。实时容错路由协议能使网络不会因为失效节点而出现传输中断、数据丢失、网络拥塞等现象,是实现网络可信传输的基础。现有实时容错协议对可能存在的拥塞造成的影响估计不足、缺乏较为有效的反馈机制、传输成功率低。本文提出了一种面向安全关键无线传感器网络的实时容错路由协议(Dynamic juMping Real time Fault tolerant protocol, DMRF),网络正常传输时,数据包以逐跳的形式从源节点发往Sink节点。当遭遇到网络拥塞、失效节点、空区域或者数据包剩余时间百分比小于阈值时,采用跳跃式传输数据,避免上述情况造成的传输中断现象,减少传输延迟,增加传输成功率。实验结果表明,DMRF能够有效防止失效节点、网络拥塞、空区域造成的影响,是一种低复杂度、高效率、低功耗的实时容错路由协议。(3)网络脆弱性评估与自私节点防护。安全关键无线传感器网络易遭受各类网络攻击、节点异常行为难以控制。为了减小攻击带来的破坏,常用的方法是对网络进行脆弱性评估,加强对脆弱节点的保护。然而现有的评估算法对脆弱性和攻击破坏性评估不够准确。限制节点行为方面,现有方法虽然能够检测出自私节点,但缺乏对节点自私行为的控制。本文面向安全关键无线传感器网络从网络脆弱性评估和自私节点防护两个方面确保网络的安全运行。网络脆弱性评估方面,本文提出了节点俘获攻击下静态网络和动态网络的脆弱性评估方法,从攻击者的角度分析出网络的脆弱环节,将网络的脆弱性评估转化为攻击者攻击破坏性评估。静态网络中,根据节点、链路、路径、路由间共享密钥关系,分别从攻击图、矩阵和攻击效率三个不同角度建模并评估攻击的破坏性。在动态网络中,建立连通支配集作为网络的虚拟骨架,分析攻击网络骨架中节点造成的破坏,从集中式和分布式两个角度实施攻击,以达到破坏最大化。节点自私行为限制方面,运用博弈理论设计节点的交互方式,限制节点行为,从簇内和簇间两个角度对节点行为建模并限制自私行为,防止自私节点对网络造成影响。实验表明,在脆弱性评估方面,本文攻击方法能够获得更高的攻击效率和更大的破坏性,进而更加准确地评估网络的脆弱性。自私节点防御方面,能够有效控制节点行为,使自私节点与正常节点行为无异,进而均衡网络能量消耗,延长网络生存时间。(4)基于混合聚类的无线传感器网络隐私保护算法。安全关键无线传感器网络中,一旦数据包中关于节点位置、身份等重要信息的敏感数据被截获,节点的位置隐私和数据隐私将遭到破坏。现有隐私保护方法虽能用时空匿名的方法隐藏节点的精确位置信息,但忽视了数据与位置间内在关联关系,因此为了保护节点的位置信息,需要从位置隐私和数据隐私两方面同时采取措施。本文提出了一种面向安全关键无线传感器网络的混合聚类隐私保护协议(Enhanced ClusterCloak, ECC),能够同时保护节点的位置隐私和数据隐私。位置隐私保护方面,采用K均值聚类,迭代划分区域,满足K匿名的前提下,保证区域的精确性。数据隐私保护方面,运用层次聚类的方法,防止攻击者实施关联攻击,从数据中分析出节点的真实身份。实验表明ECC能够提供节点超过预期的匿名等级,更好的保障节点的位置隐私安全,并能提供更加精确的数据信息。更多还原

【Abstract】 Safety critical wireless sensor networks usually require critical guarantees of reliability, real time and safety protection of the wireless sensor networks, which ensures effective, timely, reliable and secure data transmission. The potential applications of the safety critical systems involve industrial monitoring and control, intelligent transportation system and intelligent medical treatment, which have broad and essential application prospects.The features of the safety critical wireless sensor networks, such as dynamicity and open environment pose great challenges for designing efficient and confident protocols for wireless sensor networks. When designing protocols, it is essential to take the non-deterministic factors into consideration, such as limited resources, random topology, to overcome the drawbacks such as the unbalanced energy distribution, short network lifetime and inadequacy of safety estimation for developing the effective, real time, safe and confident protocol infrastructure of the wireless sensor networks. In this paper, we focus on studying and solving the existing problems from the perspectives of transmission efficiency, network fault tolerance, network security and privacy. The research issues can be summarized as follows.(1) Efficient data aggregation protocol based on ant colony optimization. The efficiency of the data aggregation influences the distribution of the energy cost of the network, which has a great impact on the lifetime of the network. Existing methods lack of the designing and maintaining of the dynamical topology, which leads to great consumption of the energy cost and unbalanced energy cost. In this paper, we design a class of efficient data aggregation algorithms based on the ant colony optimization for the wireless sensor networks (Data Aggregation based on Ant Colony Algorithm, DAACA), which combine the theory of the ant colony optimization with the data aggregation protocol. The pheromones in ant colony optimization are used for determining the merging node and selecting the next hop. Then the dynamical energy-aware network topology is constructed for balancing the energy cost and reducing the transmission hops. We develop the way of adjusting pheromones for adjusting the network topology, which optimizes the data aggregation tree. Based on these, we propose three heuristic algorithms, which optimize the adjustments of the pheromones from the perspective of global, threshold and ant colony system for further promoting the transmission efficiency and save costs. Experimental results reveal that, DAACA has longer lifetime, lower average energy cost and fewer transmission hops. (2) Real time fault tolerant protocol based on jumping transmission. The real time fault tolerant routing protocol guarantees the faulty nodes will never cause the network interrupt, data loss or network congestion, which is the basis of the confident transmission of the networks. Existing real time fault tolerant protocols underestimate the influences of the potential congestion, lack of feedback mechanism and have low transmission success ratio. In this paper, we propose a real time fault tolerant routing protocol (Dynamic juMping Real time Fault tolerant protocol, DMRF) for safety critical wireless sensor networks. When the network works normally, data are transmitted hop-by-hop from the source node to the sink node. If network congestion, faulty nodes, void region exist, or the remaining time ratio of the data packet is approaching the threshold, packets will be transmitted in a jumping mode to avoid these three circumstances, which reduce the transmission latency and enhance the transmission successful rate. Experimental results manifest that DMRF can effectively avoid the influences of the faulty nodes, network congestion and void region, it is a low complexity, efficient and low cost real time fault tolerant routing protocol.(3) Network vulnerability evaluation and selfish nodes avoidance. The safety critical wireless sensor networks are prone to various attacks and the behaviors of the nodes are difficult to control. To migrate the destructiveness of the attack, a common method is to use the vulnerability evaluation to enhance the security level of the vulnerable points. However, the precision of the previous methods in the vulnerability of the network and destructiveness of the attack are low. In the perspective of the selfish behavior avoidance, although, existing methods can detect the selfish nodes, the mechanisms of controlling the selfish behavior are lacked. In this paper, we focus on vulnerability evaluation and selfish nodes avoidance for the safety critical wireless sensor networks. In the vulnerability evaluation aspect, we propose the evaluation methods for static network and dynamic network under the node compromise attack. We analyze the vulnerable point from an adversarial point and transform the vulnerability evaluation to the destructiveness evaluation of the attacker. In the static network, based on the key sharing relationships between nodes, links, paths and routes, we develop modeling for estimating the destructiveness in terms of attacking graphs, matrix and energy efficiency. For the dynamical network, we construct the connected dominating set as the virtual backbone. We analyze the destructiveness when attacking the nodes in the backbone in a centralized and distributed way respectively to seek for maximum destructiveness. In the perspective of selfish behavior limitation, game theory is applied in designing the communication process between nodes. We regulate the behaviors of the nodes within clusters and among clusters, which avoids the destruction to the networks. Simulation results show that, in the vulnerability evaluation aspect, our methods owns higher attacking efficiency and more destructiveness, which indicates more precise vulnerability evaluation. In selfish behavior avoidance aspect, our methods can effectively regulate the behaviors of the network, which limits the selfish nodes’ behaviors as same as the normal nodes. This characteristic balances the energy cost of the network and prolongs the lifetime.(4) Network privacy protection algorithms based on combined clustering. In the safety critical wireless sensor networks, the location information and identity information are recorded in the data packets. Once the packets are intercepted, the location information and data information of the nodes will be destroyed. Although existing approaches use the spatial and temporal methods to conceal the exact location information of the nodes, they neglect to analyze the inner relationships between location and information. Therefore, to protect the location information, we need to take actions from both location information and data information. In our works, we develop a combined clustering algorithm (Enhanced ClusterCloak, ECC) for protecting location privacy and data privacy simultaneously. In location privacy protecting aspect, K-means clustering is used for diving regions iteratively, which meets the needs of K-anonymity while guaranteeing the region privacy. In the data privacy protection aspect, the hierarchical clustering is applied, which forbids the attacker from figuring out the real identities of the nodes. Simulation results indicate that ECC can provide nodes with higher anonymity level than expected, and it can protect the location privacy for the nodes while providing more exact data information.更多还原