Linkboost: A Link Prediction Algorithm to Solve the Problem of Network Vulnerability in Cases Involving Incomplete Information

• Main Authors: Chengfeng Jia, Jie Ma, Qi Liu, Yu Zhang, Hua Han
• Format: Article
• Language: English
• Published: Hindawi-Wiley 2020-01-01
• Series: Complexity
• Online Access: http://dx.doi.org/10.1155/2020/7348281

The vulnerability of network information systems has attracted considerable research attention in various domains including financial networks, transportation networks, and infrastructure systems. To comprehensively investigate the network vulnerability, well-designed attack strategies are necessary. However, it is difficult to formulate a global attack strategy as the complete information of the network is usually unavailable. To overcome this limitation, this paper proposes a novel prediction algorithm named Linkboost, which, by predicting the hidden edges of the network, can complement the seemingly missing but potentially existing connections of the network with limited information. The key aspect of this algorithm is that it can deal with the imbalanced class distribution present in the network data. The proposed approach was tested on several types of networks, and the experimental results indicated that the proposed algorithm can successfully enhance the destruction rate of the network even with incomplete information. Furthermore, when the proportion of the missing information is relatively small, the proposed attack strategy relying on the high degree nodes performs even better than that with complete information. This finding suggests that the nodes important to the network structure and connectivity can be more easily identified by the links added by Linkboost. Therefore, the use of Linkboost can provide useful insight into the operation guidance and design of a more effective attack strategy.