Disconnecting impaired or suspicious nodes and rewiring to those reliable, adaptive networks have the potential to inhibit cascading failures, such as DDoS attack and computer virus. The weights of disconnected links, indicating the workload of the links, can be transferred or redistributed to newly connected links to maintain network operations. Distinctively different from existing studies focused on adaptive unweighted networks, this paper presents a new mean-field model to analyze the reliability of adaptive weighted networks against cascading failures. By taking mean-field approximation, we develop a new continuous-time Markov model to capture the propagations of cascading failures and the rewiring actions that individual nodes can take to bypass failed neighbors. We analyze the stability of the model to identify the critical conditions, under which the cascading failures can be eventually inhibited or would proliferate. The conditions are evaluated under different link weight distributions and rewiring strategies. Our model reveals that preferentially disconnecting suspicious peers with high weights can effectively inhibit virus and failures.

中文翻译：

---

大规模自适应加权网络的可靠性分析


断开受损或可疑节点的连接并重新连接到那些可靠的自适应网络有可能抑制级联故障，例如 DDoS 攻击和计算机病毒。断开链路的权重指示链路的工作负载，可以将其转移或重新分配到新连接的链路以维持网络运行。与关注自适应未加权网络的现有研究截然不同，本文提出了一种新的平均场模型来分析自适应加权网络针对级联故障的可靠性。通过采用平均场近似，我们开发了一种新的连续时间马尔可夫模型来捕获级联故障的传播以及各个节点可以采取的绕过故障邻居的重新布线操作。我们分析模型的稳定性以确定临界条件，在这些条件下，级联故障最终可以被抑制或扩散。在不同的链路权重分布和重新布线策略下评估这些条件。我们的模型表明，优先断开具有高权重的可疑节点可以有效抑制病毒和故障。