|
| 引用本文： |
-
刘永继,陈藩松,祝博远,林伟诚,朱红松,孙利民.一种差异化信息传播聚合的复杂网络分析算法[J].信息安全学报,已采用 [点击复制]
- liuyongji,chenfansong,zhuboyuan,linweicheng,zhuhongsong,sunlimin.Identification of Influential Nodes in Complex Networks with Differentiated Propagation and Aggregation[J].Journal of Cyber Security,Accept [点击复制]
|
|
| |
|
|
| 本文已被：浏览 1118次 下载 15次 |
|
| 一种差异化信息传播聚合的复杂网络分析算法 |
|
刘永继1, 陈藩松2, 祝博远2, 林伟诚2, 朱红松1, 孙利民1
|
|
|
| (1.中国科学院信息工程研究所;2.中国科学院大学) |
|
| 摘要: |
| 随着信息系统的演进，计算机网络自治系统变得越来越复杂，涉及各种网络角色，这对网络安全和性能提出了新的挑战。复杂网络以其在电网和社交网络等领域的重要性而闻名，而计算机网络是一种由终端、交换机以及路由器组成的复杂网络，对计算机网络进行关键节点分析并定制安全措施有助于预测病毒传播、检测异常流量以及在优化网络性能的同时增强安全性。已有的关键节点识别方法包括基于度数的方法、基于重力模型的方法或基于图卷积网络的方法等，但他们存在识别准确率低，排名分辨率差以及普适性低等问题。受图卷积网络（GCN）的节点信息聚合的启发，我们提出了一种差分信息传递与聚合的方法KSDPA，这是一种基于K-Shell值聚合信息的新颖方法。我们的方法通过基于K-Shell考虑全局网络信息，在2R轮迭代中执行全局和局部影响力分析，其中在前R轮中按重要性对节点进行梯队划分，并在后续R轮中对每个层内的节点进行细粒度地排名。通过对16个真实网络的实验证明，我们的方法在这些网络中的13个网络中准确性表现最佳，12个网络中分辨率表现最佳。同时，我们的方法比EHCC方法在准确率方面提高了5%，并比DGCM+方法拥有更高的排名分辨率。最后我们对一个计算机网络拓扑结构进行分析并提出了我们的建议。 |
| 关键词: 复杂网络 关键节点 计算机网络 k-shell 信息聚合 |
| DOI：10.19363/J.cnki.cn10-1380/tn.2025.04.11 |
| 投稿时间：2023-11-15修订日期：2024-01-26 |
| 基金项目:国家自然科学基金项目, 中国科学院青年创新促进会 |
|
| Identification of Influential Nodes in Complex Networks with Differentiated Propagation and Aggregation |
|
liuyongji1, chenfansong2, zhuboyuan2, linweicheng2, zhuhongsong1, sunlimin1
|
| (1.Institute of Information Engineering, Chinese Academy of Sciences;2.University of Chinese Academy of Sciences) |
| Abstract: |
| As information systems continue to evolve, computer network autonomous systems are becoming increasingly complex, encompassing a diverse array of network roles. This growing complexity introduces fresh challenges concerning network security and performance. Notably, complex networks, which have demonstrated their significance in various fields like power grids and social networks, form the foundation of computer networks. The analysis of key nodes within computer networks and the customization of security measures play pivotal roles in predicting the spread of viruses, detecting ab-normal traffic patterns, and optimizing network performance. Existing methods for identifying key nodes encompass de-gree-based approaches, gravity model-based methods, and graph convolution network-inspired techniques. However, these methods often exhibit limitations, such as low identification accuracy, inadequate ranking resolution, and limited applicability. Drawing inspiration from the information aggregation techniques in graph convolutional networks (GCN), we introduce KSDPA, a novel method for information aggregation based on K-Shell values. K-Shell incorporates global net-work information, while our approach engages in both global and local impact analyses across 2R iterations. The first R rounds stratify nodes based on their importance, while the subsequent R rounds further stratify and rank nodes within each layer. Extensive experimentation conducted across 16 real networks underscores the efficacy of our approach. KSDPA excels in terms of accuracy within 13 of the 16 networks and achieves superior ranking resolution in 12 of them. In comparison to the EHCC method, our approach boosts accuracy by 5% while outperforming the DGCM+ method in ranking resolution. In conclusion, our research underscores the increasing complexity of computer network systems and the significance of identifying key nodes for network optimization and security. The KSDPA method offers a substantial improvement in accuracy and ranking resolution, showcasing its potential for enhancing network analysis and perfor-mance. Furthermore, we present a detailed analysis of a computer network topology and provide our recommendations for further research and development in this domain. |
| Key words: complex networks influential nodes computer networks k-shell information aggregation |
|
|
|
|
|
