无线传感器网络

传感器基础知识

由于超大规模集成电路 (VLSI) 以及微机电系统科技 (MEMS technology) 等硬件基础以及radio frequency (RF) 技术的进步，使得传感器的发展愈来愈快。

传感器具备的优点：

1. 能够放置在任何环境，任什么时候间均可以工做，而且不须要太多的人力来进行管理。
2. 具备更好的容错能力，局部出现故障仍然能较好的完成工做。
3. 获取的数据更精确。经过多个传感器获取的信息更加可靠准确。
4. 成本低以及容易部署。

Due to WSNs’reliability, self-organization, flexibility, and ease of deployment, their existing and potential applications vary widely. As well, they can be applied to almost any environment, especially those in which conventional wired sensor systems are impossible or unavailable, such as in inhospitable terrains, battlefields, outer space, or deep oceans.

传感器用于通讯耗费的资源比用于感知和计算花费的资源多[1,12]：It is reported that the energy consumed by communication is much higher than that for sensing and computation; in fact, this actually dominates the total energy consumption in WSNs. Furthermore, in most WSNs, power for transmission contributes to a majority of the total energy consumed for communication and the required transmission power grows exponentially with the increase of transmission distance. Therefore, reducing the amount of traffic and distance of communications can greatly prolong the system’s lifetime.

无线传感器网络与传统无线自组织网络的区别[1]

无线传感器网络的分类

由于不一样的传感器使用的环境不一样，相应的技术要求也不一样，所以传感器经常是面向应用来设计的，因此不一样无线传感器网络的架构 (architectures)、协议 (protocols) 以及算法也每每不同。虽然如此，不一样的传感器网络仍然具备一些共同的特色，通常有如下几种划分方式[1]：

1. 根据WSNs节点到基站的距离，能够划分为单跳 (single-hop, also known as nonpropagating) 和多跳 (multihop or propagating) 系统。在单跳系统中，全部的传感器节点直接把数据传输给基站，在多跳系统中数据须要通过中间传感器节点才传到基站。较小的传输功率可能会因为跳数增长和再传输几率增大而致使功率消耗过大[9]，不过文献[10]中提供了一种方法能够对多跳系统传输的能量进行估计。Single-hop networks have much simpler structure and control and fit into the applications of small sensing areas; multihop networks promise wider applications at the cost of higher complexity.

更多有关单跳和多跳系统[1]：

As reported in Rappoport [7], large-scale propagation follows as exponential law to the transmitting distance (usually with exponent 2 to 4 depending on the transmission environment). It is not difficult to show that power consumption due to signal transmission can be saved in orders of magnitude by using multihop routing with short distance of each hop instead of single-hop routing with a long range of distance for the same destination. And the majority of existing WSN literature is based on multihop ad hoc architectures.

1. 基于传感器节点密度和数据依赖性，WSNs能够划分为聚合的和非聚合的网络。在非聚合系统中，来自每一个节点的全部数据将被发送到目的地，中间节点的计算负载相对较小，系统能够达到高精度。可是，随着网络规模的扩大，整个系统的总流量负载可能会迅速增长，通讯将消耗更多的能量，会产生更多的冲突和/或拥塞，致使高延迟。Therefore, the nonaggregating scheme is suitable for systems that have less node density, sufficient capacity, and/or in which extremely high accuracy is
   demanded by end users. 而在密集分布的网络中，传感器节点一般位于邻近节点附近。所以，来自多个源的信息可能高度相关，而且能够在中间节点上执行聚合函数以消除数据冗余。这样，系统的交通负荷就会大大下降，而且能够得到大量的通讯能量。可是，中间节点将执行计算函数，这可能须要更大的内存。
2. 根据传感器节点的分布，WSNs能够是肯定性的或动态的。在肯定性系统中，传感器节点的位置是固定的或预先计划的。该系统的控制更简单，实现也更容易。可是，该方案仅适用于可以提早获取和规划传感器节点位置信息的有限系统。然而，在许多状况下，传感器节点的位置是不可提早得到的，例如在偏远地区随机丢弃的节点。动态方案具备更强的可扩展性和灵活性，但须要更复杂的控制算法。
3. 根据控制方案，WSNs能够是不可自我配置的，也能够是自我配置的。在前一种机制中，传感器节点不能自行组织，而是依靠中央控制器向它们提供命令并收集信息。该方案只能用于小型网络。然而，在大多数WSNs中，传感器节点能够自行创建和维护链接，协同完成传感和控制任务。这种自配置方案更适合大型系统执行复杂的监视任务和信息收集和传播。尽管自我配置系统比非自我配置系统更复杂，但在实际应用中更实用，特别是当网络规模变得很是大时。然而，它们提出了许多挑战和有待进一步探讨的开放问题。
   总结见下表[1]。
   

性能评估方法

无线传感器网络是一种特殊的自组织网络（Ad hoc networks），和自组织网络同样面临着能量有限（energy constraints）以及路由选择（routing）的挑战，其中能量有限（energy constraints）是无线传感器最大的挑战。

有不少指标能够用来评估传感器网络的性能状况，其中主要有：

1. Energy efficiency/system lifetime. The sensors are battery operated, rendering energy a very scarce resource that must be wisely managed in order to extend the lifetime of the network[2].
2. Latency. Many sensor applications require delay-guaranteed service. Protocols must ensure that sensed data will be delivered to the user within a certain delay. Prominent examples in this class of networks are certainly the sensor-actuator networks.
3. Fault tolerance. Robustness to sensor and link failures must be achieved through redundancy and collaborative processing and communication.
4. Scalability. Because a sensor network may contain thousands of nodes, scalability is a critical factor that guarantees that the network performance does not significantly degrade as the network size (or node density) increases.
5. Transport capacity/throughput. Because most sensor data must be delivered to a single base station or fusion center, a critical area in the sensor network exists, whose sensor nodes must relay the data generated by virtually all nodes in the network. Thus, the traffic load at those critical nodes is heavy, even when the average traffic rate is low. Apparently, this area has a paramount influence on system lifetime, packet end-to-end delay, and scalability.

提升传输效率的方法

能够利用下面几种方式来提升传感器的能量利用效率[1]：

• 首先，能够在集群中应用分布式源编码的形式进行数据压缩，以减小要传输的数据包数量。First, data compression in the form of distributed source coding is applied within a cluster to reduce the number of packets to be transmitted.
• 其次，数据中心化的属性使传感器节点的标识 (例如地址) 变得用处不大，然而在不少场景中咱们比较关注某些区域的数据特色，这时就要去除中心化，利用分布式处理数据的方法。Second, the data-centric property makes an identity (e.g., an address) for a sensor node obsolete. In fact, the user is often interested in phenomena occurring in a specified area[3], rather than in an individual sensor node.
• 此外，也能够利用广播树和多播树策略来提升能量利用效率（该方法缺点：会致使较高的计算复杂度）。Another strategy to increase energy efficiency is to use broadcast and multicast trees.
• 最后，引入传感器睡眠机制减小电量消耗[11]。The exploitation of sleep modes is imperative to prevent sensor nodes from wasting energy in receiving packets unintended for them. For example, one option is to use coordinated policy for deciding which node will go to sleep while the connectivity in the node is maintained[15].

建模

通常对传感器建模时考虑传感器可能处于四种状态：传输信号(transmission), 信号检测(reception), 信号接收(listening), 睡眠(sleeping)。其中，

• Transmission: processing for address determination, packetization, encoding, framing, and maybe queuing; supply for the baseband and RF circuitry.
• Reception: Low-noise amplifier, downconverter oscillator, filtering, detection, decoding, error detection, and address check; reception even if a node is not the intended receiver.
• Listening: Similar to reception except that the signal processing chain stops at the detection.
• Sleeping: Power supply to stay alive.

传感器网络存在的技术挑战和解决方法

技术挑战

目前主要的挑战问题[1]：

1. Massive and random deployment. Most WSNs contain a large number of sensor nodes (hundreds to thousands or even more), which might be spread randomly over the intended areas or are dropped densely in inaccessible terrains or hazardous regions. The system must execute self-configuration before the normal sensing routine can take off.
2. Data redundancy. The dense deployment of sensor nodes leads to high correlation of the data sensed by the nodes in the neighborhood.
3. Limited resources. WSN design and implementation are constrained by four types of resources: energy, computation, memory, and bandwidth. Constrained by the limited physical size, microsensors could only be attached with bounded battery energy supply. Moreover, WSNs usually operate in an untethered manner, so their batteries are nonrechargeable and/or irreplaceable. At the same time, their memories are limited and can perform only restricted computational functionality.
   The bandwidth in the wireless medium is significantly low as well.
4. Ad hoc architecture and unattended operation. The attributes of no fixed infrastructure and human-unattended operation of such networks require the system to establish connections and maintain connectivity autonomously.
5. Dynamic topologies and environment. On the one hand, the topology and connectivity of WSNs may frequently vary due to the unreliability of the individual wireless microsensors. For example, a node may fail to function because of exhaustion of power at any time without notification to other nodes in advance. As well, new nodes may be added randomly in an area without prior notification of existing nodes. On the other hand, the environment that the WSNs are monitoring can also change dramatically, which may cause a portion of sensor nodes to malfunction or render the information they gather obsolete.
6. Error-prone wireless medium. Sensor nodes are linked by the wireless medium, which incurs more errors than their wired counterpart. In some applications, the communication environment is actually noisy and can cause severe signal attenuation.
7. Diverse applications. WSNs could be used to perform various tasks, such as target detection and tracking, environment monitoring, remote sensing, military surveillance, etc. Requirements for the different applications may vary significantly.
8. Safety and privacy. Safety and privacy should be an essential consideration in the design of WSNs because many of them are used for military or surveillance purposes. Denial of service attacks against these networks may cause severe damage to the function of WSNs. However, security seems to be a significantly difficult problem to solve in WSNs because of the inevitable dilemma: WSNs are resource limited and security solutions are resource hungry. Indeed, most existing communication protocols for WSNs do not address security and are susceptible to adversaries [9].
9. QoS concerns. The quality provided by WSNs refers to the accuracy with which the data reported match what is actually occurring in their environment. Different from others, accuracy in WSNs emphasizes the characteristic of the aggregated data of all sources instead of individual flows. One way to measure accuracy is the amount of data. Another aspect of QoS is latency. Data collected by WSNs are typically time sensitive, e.g., early warning of fires. It is therefore important to receive the data at the destination/control center in a timely manner. Data with long latency due to processing or communication may be outdated and lead to wrong reactions.

传感器的协做问题

由于有时候须要对一个移动的目标进行数据采集，好比说移动目标地理位置信息，不少时候要想准确获得这样的信息须要多种不一样传感器之间协做，由于单个传感器采集的数据有时会出错以及误报。如今能够采用分布式计算技术来解决这一问题[1]：

These capabilities are now being extended to include high-speed wireless and fiber networking with distributed computing. As the Internet protocol (IP) technologies continue to advance in the commercial sector, the military can begin to leverage IP formatted sensor data to be compatible with commercial high-speed routers and switches. Sensor data from theater can be posted to high-speed networks, wireless and fiber, to request computing services as they become available on this network. The sensor data are processed in a distributed fashion across the network, thereby providing a larger pool of resources in real time to meet stringent latency requirements. The availability of distributed processing in a grid-computing architecture offers a high degree of robustness throughout the network. One important application to benefit from these advances is the ability to geolocate and identify mobile targets accurately from multiaspect sensor data.

目前协做存在的不足：The limitation is with the communication and available distributed computing.

多跳决策问题

无线传感器网络一般在很宽的区域内包含大量的传感器节点，基站可能远离无线传感器。所以，将整个系统划分红不一样的集群，用多跳短距离数据转发代替单跳远程传输。这将减小数据通讯消耗的能量，而且在网络规模增加时具备负载平衡和可伸缩性的优点。这种基于集群的方法面临的挑战包括如何选择集群头以及如何组织集群。有大量文章介绍了如何选择集群头，选择集群头要注意的一个问题是：不只要保证WSNs效率高，并且各个集群头的负载要平衡（见文献[1]7-16）

去除冗余信息问题

在传感器通讯过程当中可能存在冗余信息，去除多余冗余信息将显著提升效率。The most straightforward is duplicate suppression, i.e., if multiple sources send the same data, the intermediate node will only forward one of them. Maximum or minimum functions are also very simple approaches. Heinzelman and colleagues [13] and Julik and coworkers [14] propose a scheme named sensor protocols for information via negotiation (SPIN) to realize traffic reduction for information dissemination. It introduces metadata negotiations between sensors to avoid redundant and/or unnecessary data through the network.

能量补充问题

能量问题是传感器网络须要解决的重要问题。能源供应至少能够用两种概念上不一样的方式来解决，第一种方式是为每一个传感器节点配置一个 (可再充电) 的能量源，能够经过两种途径: (1)选择是使用高密度电池 (当前主要方式); (2)使用燃料电池 (full cells)，但目前还不能很好地应用于传感器上；第二种方式是选择在环境中获取能源，好比说太阳能电池、热能电池。

定位问题

WSNs的部署常常是随机放置的，与之相关的主要挑战之一是选择传感器的类型和数量并肯定它们的位置。这一任务是困难的，由于有许多类型的传感器具备不一样的属性，如分辨率、成本、精度、大小和功耗。不过这一问题有解决方法：For example, consider determining distance using audio sensors. Because the speed of sound depends greatly on temperature and humidity of the environment, it is necessary to take both measurements into account in order to get the accurate distance. 已有的传感器定位技术有:VM[17], SeRLoc[18].

传感器网络的管理

由于绝大部分传感器比较廉价，使用时投放量多，范围广，而且比较隐蔽，这使得若是去维护传感器每每成本较高，并且效率很低，因此每每传感器一经投放就几乎不多去维护，这些特色使得传感器必须具有自治能力，也即自我管理能力 (self-managed, including self-organizing, self-healing, self-optimizing, self-protecting, self-sustaining, self-diagnostic) ，这就是咱们称为传感器网络为自组织网络的缘由。

A managed WSN with this has various characteristics can be called an autonomic system[4], which is an approach to self-managed computing systems with a minimum of human interference. The processors in such systems use algorithms to determine the most efficient and cost-effective way to distribute tasks and store data. Along with software probes and configuration controls, computer systems will be able to monitor, tweak, and even repair themselves without requiring technology staff — at least, that is the goal.

传感器网络自我管理存在的问题

通常传统的计算机网络在设计和部署时考虑到了便于管理员来维护的因素，然而传感器网络自己就不多考虑到维护问题，因此传感器的管理每每是指自我管理，不一样于传统计算机网络的由管理员来管理的特色。对传感器网络来讲，由于传感器能量有限，因此传感器网络的全部执行的操做必需要求是高效节能的。此外，传感器网络在工做过程当中可能会发生故障或者能量耗尽，这致使了网络的拓扑结构是是不断动态退化的。

传感器网络自我管理的主要体现的方面[1]：

A managed WSN is responsible for configuring and reconfiguring under varying (and, in the future, even unpredictable) conditions. System configuration (“node setup” and “network boot up”) must occur automatically; dynamic adjustments need to be done to the current configuration to best handle changes in the environment and itself. A managed WSN always looks for ways to optimize its functioning; it will monitor its constituent parts and fine-tune workflow to achieve predetermined system goals. It must perform something akin to healing — it must be able to recover from routine and extraordinary events that might cause some of its parts to malfunction. The network must be able to discover problems or potential problems, such as uncovered area, and then find an alternate way of using resources or reconfiguring the system to keep it functioning smoothly. In addition, it must detect, identify, and protect itself against various types of attacks to maintain overall system security and integrity. A managed WSN must know its environment and the context surrounding its activity and act accordingly. The management entities must find and generate rules to perform the best management of the current state of the network.

服务管理

传感器的服务和许多应用软件的功能模块有关，传感器基本的服务包括 感知(sensing)、数据处理(processing)，以及 数据分发(data dissemination)[5]。传感器管理主要有两个方面：quality of service (QoS) and denial of service (DoS).

其中，涉及到对WSNs的QoS支持的组成部分主要包括 QoS models，QoS sensing，processing 以及 QoS dissemination [6]. The larger the number of monitored QoS parameters is, the larger the energy consumption and the lower the network lifetime are.

1. QoS model. A QoS model specifies an architecture in which some of the services can be provided in WSNs.
2. QoS sensing. QoS sensing considers the sensor device calibration, environment interference monitoring, and exposure (time, distance, and angle between sensor device and phenomenon).
3. QoS dissemination. Reliable data delivery is still an open issue in the context of WSNs. QoS dissemination in WSNs is a challenging task because of constraints, mainly energy and dynamic topology of WSNs.
4. QoS processing. Processing quality depends on the robustness and complexity of the algorithms used, as well as processor and memory capacities. The way to measure processing performance changes from processor speed to the immediacy and accuracy of the response and energy consumption.

传感器网络的安全

目前的研究主要集中在提供最节能的路由。在无线传感器网络中，安全有效的路由协议和高效的路由协议做为攻击是很是须要的。在无线传感器网络中，须要安全和高效的路由协议，好比地陷 (sinkhole)、虫洞 (wormhole) 和 Sybil 攻击。另外，无线传感器网络中，数据包传输经常会遇到丢包 (missing packets)、伪造和篡改、冲突 (Conflicts)、延迟 (Latency)、非法操做 (illegal operation) 等。

虫洞攻击是指恶意节点窃听一个数据包或一系列数据包，经过传感器网络将其传输到另外一个恶意节点，而后从新播放数据包。

有关安全的保护技术

三种密码学方法已经被应用于现实系统的安全中：防火墙 (firewalls), 蜜罐技术 (honeypots), 入侵检测技术 (intrusion detection techniques). 介绍以下[1]：

1. 防火墙：A firewall is a policy enforcement point (node) for a part of a network designed to restrict access from and to that subnetwork. Several classes of firewalls exist: packet filtering according to a particular set of rules; access to particular servers or ports; or application-level firewalls that protect by remembering the state of the network connection. Firewalls still face denial of service (DoS) attacks and they try to address them by filtering suspicious connections. Among the several limitations of firewalls is the fact that they do not protect the network from insider attacks and that filtering can only be done against already known attacks.
2. 蜜罐技术：Honeypots are systems placed on networks specifically for the purpose of being attacked or compromised. Because they are not designed for true use, they exist only to detect and collect information about security attacks. Advantages of honeypots include low false positives; ability to capture unknown attacks; and ability to facilitate interaction with the attacker in order to gain better insights into actions and thinking. Intrusion detection techniques aim at recognizing statistical or pattern irregularities in the incoming or outgoing traffic. The most recent approach to detection of Internet attacks is probabilistic deduction of the IP traceback. Finally, virtual private networks are logical extensions of private networks over insecure channels provided by the Internet.

因为传统的密钥交换技术使用非对称密码技术，也称为公钥密码技术。在无线传感器网络中，非对称密码技术的问题在于它对于传感器网络中的各个节点而言一般计算量太大，于是很难适用。不过，对称密码技术须要耗费的计算量小不少，能够应用于传感器网络，可是对称密码技术安全的前提是双方共享一个密钥，而且在一方给另外一方发送密钥时没有被监听，那么双方通讯是安全的。然而，实际中很难保证不被监听，因此对称加密相对容易被破解，所以会带来安全问题。

由于在传感器网络中，远程配置和应用程序代码更新须要经过移动代码的注入以及传播来完成。合法的移动代码经过几个节点注入到网络中，而后在网络中传播[16]。保护移动代码安全的方法有：代码签名 (code signing), 沙盒 (sand-boxes), 以及 携带证实的代码 (proof-carrying code).

常见网络攻击类型

WSNs一旦部署，恶意攻击对WSN中用于管理和代码更新的节点的访问将产生安全威胁并消耗资源。尽管避免攻击很困难，但容许在应用程序和系统中修改节点的机制是必要的。在移动代码入侵技术中，最流行的有四种: 病毒(Virues)、特洛伊木马(Trojan horses)、缓冲区溢出(buffer overflow) 和 秘密通讯通道(covert communication channels)。

隐蔽通讯信道是由计算机系统中的资源共享引发的。例如，具备高优先级的进程能够经过干扰或避免干扰进程的时间来将信息传递给具备低优先级的进程。

几种经常使用攻击方法

攻击能够经过多种方式执行，这里介绍几种[20]，最明显的是拒绝服务攻击，此外还有流量分析，隐私侵犯，物理攻击等等。对无线传感器网络的拒绝服务攻击的范围能够从简单地干扰传感器的通讯信道到旨在违反802.11 MAC协议或任何其余无线传感器网络层的更复杂的攻击[19]。致使网络拒绝服务的方式有不少，其中网络拥塞是最多见的一种，也即网络中部分信道或者部分节点过载而没法正常工做，对于拥塞致使的拒绝服务攻击，经常使用的解决策略是判断出拥塞所在的位置后，利用路由绕过拥塞的部分。

除了拒绝服务攻击外，还有女巫攻击 (Sybil Attacks) [21]: Sybil攻击被定义为“非法采起多重身份的恶意盗窃行为”。它最初被描述为可以击败对等网络中分布式数据存储系统冗余机制的攻击。除了战胜分布式数据存储系统以外，Sybil攻击还能够有效地对抗路由算法，数据聚合，投票，公平资源分配以及阻止不当行为检测。发现女巫攻击的方法有两种，第一种是: 在无线电测试中，一个节点为它的每一个邻居分配一个不一样的信道，以便进行通讯。而后节点随机选择一个通道并侦听。若是节点被检测到通道上的传输，则假定在通道上传输的节点是物理节点(正常节点)。相似地，若是节点没有被检测到在指定通道上的传输，则该节点假定分配给该通道的标识不是物理标识(虚假节点)。第二种是: 使用随机密钥预分发技术，假设有限数量的钥匙在一个密匙环,一个节点随机生成的身份不会拥有足够的钥匙承担多重身份，所以没法在网络上交换消息由于无效的身份将没法进行加密或解密消息，或者说一个节点每生成一个新的身份就给予必定量的惩罚, 恶意节点为了减小损失而选择不改变身份。

流量分析攻击 (Traffic Analysis Attacks): 攻击者经过分析基站周围流量状况来肯定攻击目标。由于传感器网络中存在着一些计算能力比较强的节点，这些节点用来收集周围传感器的数据后进行分析和处理，具有这样能力的节点被称为“基站”，通常来讲基站的计算能力和防御能力比较强，攻击者不太可能可能花费较大成本去攻击基站，可是攻击者会选择攻击基站周围的传感器。首先，攻击者经过分析传感器传输的数据获取获得与基站有关的信息，对于一些离基站比较近的传感器节点，这些节点因为与基站通讯比较密切，产生的流量比较大，因此经常成为被攻击的目标，一旦攻击者攻陷这些传感器后会选择禁用基站或者篡改数据，使得基站收不到收据或者收到虚假数据。处理这类攻击的方式有: 伪造假数据包传输，并且尽可能保证整个网络流量的均衡，使得攻击者没法经过分析流量来选择攻击目标。

节点复制攻击 (Node Replication Attacks)：攻击者经过复制现有传感器节点的节点ID来尝试向现有传感器网络中添加节点。以此方式复制的节点可能严重中断传感器网络的性能：数据包可能被破坏，甚至发生错误路由。这些破坏可能致使网络断开，传感器读数错误等。若是攻击者经过复制节点可以得到对整个网络的物理访问权限，那么他能够将密钥复制到复制的传感器，也能够将复制的节点插入网络中的战略点。经过在特定网络中插入复制节点，攻击者能够轻松操纵网络的特定部分，并且还有可能使传感器网络彻底断开。

隐私攻击[22-23]: 传感器能够收集目标区域的信息，但是一旦传感器被攻陷，这些敏感信息就会被泄露，会致使隐私泄露问题。

物理攻击: 传感器网络一般在恶劣的室外环境中运行。在这样的环境中，传感器的小形状因子以及部署的无分散和分布式特性使得它们很是容易受到物理攻击，即因为物理节点破坏而致使的威胁。物理攻击永久破坏传感器，因此损失是不可逆转的。例如，攻击者能够提取加密秘密，篡改相关电路，修改传感器中的程序，或者在攻击者控制下用恶意传感器代替它们。

不幸的是，无线传感器网络没法承担实施许多典型防护策略所需的计算开销。

安全的广播和多播

无线传感器网络的主要通讯模式是广播 (broadcasting) 和多播 (multicasting)，例如1对N，N对1和M对N方式，而不是采用传统互联网的点对点方式通讯。

将来须要解决的问题

传感器相关研究须要在如下几方面进行拓展[1]:

1. Protocols and algorithms for WSNs with heterogeneous sensor nodes. Currently, many WSN protocols/algorithms are based on homogeneous sensor networks. However, sensors with different power capacities, sensing and transmitting range, and computing/processing abilities are usually more practical for constructing highly reliable networks.
2. Security issues. Most existing WSN communication protocols have not addressed security and are susceptible to attacks by adversaries. The issue of integrating security at the design stage in a resources-constrained WSN is a serious technical challenge.
3. Analytical modeling. More accurate and expeditious implementation of WSNs in the real world is highly dependent on the ability to devise analytical models to evaluate and predict WSNs' performance characteristics, such as efficiency for information gathering, delay properties, granularity, and energy consumption. However, due to the diverse forms of applications and massive number of nodes in a single network, many technical problems remain to be solved in modeling the behavior of WSNs.
4. Other issues. Optimal sensor node selection and allocation, discovery, localization, and network diagnoses are other open issues in this direction. Many software issues remain open as well. These include the design of distributed control and coordination algorithms to ensure balanced load assignment and energy consumption; efficient techniques for sensor data storage; and protocols with mobility consideration and dynamic group communications.

[1] Mahgoub I, Ilyas M. Smart dust : sensor network applications, architecture, and design[J]. Journal of Strain Analysis for Engineering Design, 2006, 37(1):21-31.
[2] Ephremides A. Energy concerns in wireless networks[J]. Wireless Communications IEEE, 2002, 9(4):48-59.
[3] Intanagonwiwat C, Govindan R, Estrin D. Directed diffusion:a scalable and robust communication paradigm for sensor networks[C]// ACM, 2000:56-67.
[4] Autonomic computing. https://en.wikipedia.org/wiki/Autonomic_computing. [5] Ruiz L B, Nogueira J M, Loureiro A A F. Sensor network management[M]//Handbook of Sensor Networks. CRC Press, 2004: 64-98. [6] Ruiz L B, Nogueira J M, Loureiro A A F. Manna: A management architecture for wireless sensor networks[J]. IEEE communications Magazine, 2003, 41(2): 116-125. [7] Rong Z, Rappaport T S. Wireless communications: Principles and practice, solutions manual[M]. Prentice Hall, 1996. [8] Wood A D, Stankovic J A. Denial of service in sensor networks[J]. computer, 2002, 35(10): 54-62. [9] Bansal S, Gupta R, Shorey R, et al. Energy efficiency and throughput for TCP traffic in multi-hop wireless networks[C]//INFOCOM 2002. IEEE, 2002, 1: 210-219. [10] Yeh C H. ROAD: A variable-radius MAC protocol for ad hoc wireless networks[C]//Vehicular Technology Conference, 2002. VTC Spring 2002. IEEE, 2002, 1: 399-403. [11] Raghunathan V, Schurgers C, Park S, et al. Energy-aware wireless microsensor networks[J]. IEEE Signal processing magazine, 2002, 19(2): 40-50. [12] Zhao F, Liu J, Liu J, et al. Collaborative signal and information processing: an information-directed approach[J]. Proceedings of the IEEE, 2003, 91(8): 1199-1209. [13] Heinzelman W R, Kulik J, Balakrishnan H. Adaptive protocols for information dissemination in wireless sensor networks[C]. ACM, 1999: 174-185. [14] Kulik J, Heinzelman W, Balakrishnan H. Negotiation-based protocols for disseminating information in wireless sensor networks[J]. Wireless networks, 2002, 8(2/3): 169-185. [15] Rozovsky R, Kumar P R. SEEDEX: A MAC protocol for ad hoc networks[C] ACM, 2001: 67-75. [16] Boulis A, Srivastava M B. A framework for efficient and programmable sensor networks[C] IEEE, 2002: 117-128. [17] Capkun S, Hubaux J P. Secure positioning in wireless networks[J]. IEEE Journal on Selected Areas in Communications (JSAC), 2006, 24: 221-232. [18] Lazos L, Poovendran R. SeRLoc: Robust localization for wireless sensor networks[J]. ACM Transactions on Sensor Networks, 2005, 1(1): 73-100. [19] Perrig A, Stankovic J, Wagner D. Security in wireless sensor networks[J]. Communications of the ACM, 2004, 47(6): 53-57. [20] Walters J P, Liang Z, Shi W, et al. Wireless sensor network security: A survey[J]. IN DISTRIBUTED, GRID, AND PERVASIVE COMPUTING, YANG XIAO (EDS, 2007(2):0--849. [21] Newsome J, Shi E, Song D, et al. The sybil attack in sensor networks: analysis & defenses[C] ACM, 2004: 259-268. [22] Chan H, Perrig A. Security and privacy in sensor networks[J]. computer, 2003, 36(10): 103-105. [23] Gruteser M, Schelle G, Jain A, et al. Privacy-Aware Location Sensor Networks[C] HotOS. 2003, 3: 163-168.