Luo Song, Lai Lianghai, Hu Tan, Hu Xin
College of Computer Science and Engineering, Chongqing University of Technology, Chongqing, China.
PeerJ Comput Sci. 2024 Jun 12;10:e2108. doi: 10.7717/peerj-cs.2108. eCollection 2024.
With the development of technology, more and more devices are connected to the Internet. According to statistics, Internet of Things (IoT) devices have reached tens of billions of units, which forms a massive Internet of Things system. Social Internet of Things (SIoT) is an essential extension of the IoT system. Because of the heterogeneity present in the SIoT system and the limited resources available, it is facing increasing security issues, which hinders the interaction of SIoT information. Consortium chain combined with the trust problem in SIoT systems has gradually become an important goal to improve the security of SIoT data interaction. Detection of malicious nodes is one of the key points to solve the trust problem. In this article, we focus on the consortium chain network. According to the information characteristics of nodes on the consortium chain, it can be analyzed that the SIoT malicious node detection combined with the consortium chain network should have the privacy protection, subjectivity, uncertainty, lightweight, dynamic timeliness and so on. In response to the features above and the concerns of existing malicious node detection methods, we propose an algorithm based on inter-block delay. We employ unsupervised clustering algorithms, including K-means and DBSCAN, to analyze and compare the data set intercepted from the consortium chain. The results indicate that DBSCAN exhibits the best clustering performance. Finally, we transmit the acquired data onto the chain. We conclude that the proposed algorithm is highly effective in detecting malicious nodes on the combination of SIoT and consortium chain networks.
随着技术的发展，越来越多的设备接入互联网。据统计，物联网（IoT）设备已达数百亿台，形成了庞大的物联网系统。社交物联网（SIoT）是物联网系统的重要延伸。由于SIoT系统中存在异构性且可用资源有限，它面临着日益严重的安全问题，这阻碍了SIoT信息的交互。联盟链结合SIoT系统中的信任问题，逐渐成为提高SIoT数据交互安全性的重要目标。恶意节点检测是解决信任问题的关键要点之一。在本文中，我们聚焦于联盟链网络。根据联盟链上节点的信息特征，可以分析出结合联盟链网络的SIoT恶意节点检测应具备隐私保护、主观性、不确定性、轻量级、动态及时性等特点。针对上述特点以及现有恶意节点检测方法的不足，我们提出了一种基于块间延迟的算法。我们采用无监督聚类算法，包括K均值和DBSCAN，对从联盟链截获的数据集进行分析和比较。结果表明，DBSCAN表现出最佳的聚类性能。最后，我们将获取的数据上传至链上。我们得出结论，所提出的算法在检测SIoT与联盟链网络组合中的恶意节点方面非常有效。
