中文摘要：

无线自组织网络的报文传输是依靠网络中的节点彼此多跳接力传输,当网络中有节点被俘获以后,就会成为内部攻击者,并在报文的传输过程中发起丢弃报文攻击,严重降低网络性能.现有网络协议栈中传输层和网络层协议难以检测和防范在网络层发起丢弃报文攻击的节点.文中提出一种在网络层抵抗内部节点丢弃报文攻击的通信模型,它包括通信链路状态实时分析协议和分布式的节点类型判定算法两个部分.通信链路状态实时分析协议利用节点对间逐段生成的路径环路,通过引入报文成组应答机制使得节点能实时地获取其邻居节点的报文转发状态;在通信链路状态实时分析协议基础之上,节点根据相关数学模型能有效地分析邻居节点行为是否异常,并对网络中的节点进行分类,最终将恶意节点从网络中隔离出去.仿真实验结果表明,文中算法在恶意节点的检测率和误检率方面性能表现良好,能有效地抵抗来自网络内部节点的丢弃报文攻击.

英文摘要：

In wireless ad hoc networks, packets are delivered by multi-hop relay among nodes. When some nodes are captured, they turn out to be inside attackers, then drop data packets arbi- trarily. Such kind of attack, called packet dropping attack, will dramatically degrade network performance. However, neither transport layer nor network layer protocols of existing network protocol stacks are able to defend against it efficiently. Hence, this paper presents a new secure communication model defending against packet dropping attack from inside attacker in the net- work layer, including Real-time Link Status Analysis protocol （RLSA） and Distributed Node Classification algorithm （DNC）. RLSA took advantage of segment-based multiple paths between a pair of source and destination, as well as group acknowledgement mechanism for packets, to obtain packet forwarding status of neighbors in real time. Based on RLSA and related analytical model, DNC identifies abnormal behavior of nodes, and then classifies nodes into normal and ma- licious types. Simulation results show the secure communication model performs well regarding malicious node detection rate and false alarm rate.