Mathematical epidemiological analysis of dynamics of delay attacks on pull-based competitive information diffusion
Mathematical epidemiological analysis of dynamics of delay attacks on pull-based competitive information diffusion
复制标题
DOI：
10.1016/j.comnet.2020.107383
复制
发表时间：
2020-10
期刊：
影响因子：
--
通讯作者：
Masahiro Sasabe
中科院分区：
文献类型：
--
作者：
Masahiro Sasabe
近年来，竞争信息扩散在特定的网络系统中扮演着重要的角色。例如，在加密货币系统中，比特币，一些特殊的节点，称为矿工，在扩散竞赛中相互竞争以获得系统奖励。如果用于扩散的信息很大，则通常使用基于拉的传播来减轻通信开销，其中节点在传输信息本身之前向其邻居通知小的库存消息。在接收到库存消息时，相邻节点向该节点请求相应的信息。由于信息传输可能由于临时网络/节点故障而需要很多时间，因此基于拉的传播通常涉及超时机制。如果发生超时，则节点重试信息检索。恶意节点可以利用这种超时机制对信息传播进行延迟攻击，即，在接收到请求之后暂时中断所述信息传送。本文研究了具有中断的信息扩散风险，在这种风险下，与特定源节点合谋的对手同时进行延迟攻击，以减缓竞争源节点的信息扩散。借助数学流行病学中传染病传播的确定性非线性模型，首次建立了一个标量SIRA模型，该模型描述了混合种群下具有中断的信息传播.在此基础上，通过扩展标量SIRA模型，提出了一种网络SIRA模型，该模型支持任意网络上的中断信息扩散。通过数值评估，从攻击规模、攻击率、恢复率和网络结构的角度定量地揭示了中断信息扩散的风险。
In recent years, competitive information diffusion plays a key role in specific network systems. For example, in the cryptocurrency systems, e.g., Bitcoin, some special nodes, called miners, compete with each other in diffusion races to acquire the system rewards. If the information for diffusion is large, pull-based dissemination is often used to alleviate the communication overhead, where a node notifies a small inventory message to its neighbor(s) before transferring the information itself. On receiving the inventory message, the neighboring node requests the corresponding information from the node. Since the information transfer may require much time due to temporal network/node trouble, the pull-based dissemination generally involves a timeout mechanism. If a timeout occurs, the node retries the information retrieval. A malicious node can conduct a delay attack on the information propagation by exploiting this timeout mechanism, i.e., temporarily interrupting the information transfer after receiving a request. In this paper, a risk of information diffusion with interruption is considered, under which adversaries colluding with a specific source node simultaneously conduct the delay attacks to slow down the information diffusion from competitive source nodes. With the help of the deterministic nonlinear model for the propagation of infectious diseases in mathematical epidemiology, a scalar standby-interrupted-retrieved-attackable (SIRA) model is first developed, which captures the information diffusion with interruption under a well-mixed population. Furthermore, a network SIRA model is proposed by extending the scalar SIRA model to support the information diffusion with interruption over arbitrary networks. Through numerical evaluations, the risk of information diffusion with interruption is quantitatively revealed from the viewpoint of attack scale, attack rate, recovery rate, and network structure.