网络结构鲁棒性指标及应用研究

为了更好地测度网络抵御破坏的能力,基于网络连通和恢复能力提出了连接鲁棒性和恢复鲁棒性两种指标.运用这两种指标,以网络规模为500,取20次独立实验的均值,对ER随机网络、规则网络、BA无标度网络以及WS小世界网络4种典型网络结构进行仿真.实验结果表明:ER随机网络对于恶意攻击的鲁棒性要优于其他3种网络;BA无标度网络仅节点恢复鲁棒性较好,边恢复鲁棒性和连接鲁棒性最差;规则网络拥有很好的连接鲁棒性但恢复鲁棒性最差;WS小世界网络受其参数影响,鲁棒性介于ER随机网络和规则网络之间.同时还发现,网络结构鲁棒性的下降随着去除节点个数的增加和网络结构参数的改变而呈现出一定的"涌现"现象.

Study on Indices of Network Structure Robustness and Their Application

Connectivity robustness and recovery robustness are proposed to measure the degree of tolerance against failure and attacks in networks. Simulations are carried out in four types of classic networks (i. e. , ER random network, regular network, BA scale-free network, and WS small-world network) with 500 nodes. Results obtained from an average of 20 tests on ER random network, regular network, BA scale-free network, and WS small-world network show that, ER random networks and regular networks are more robust than BA scale-free networks when involving in hostile attacks, and that WS small-world network are more robust than BA scale -free networks but less robust than random networks and regular networks. It is also found that an e-mergent decline of structural robustness appears with removing more nodes and changing network structure parameters.