With the rapid growth of Internet industry, online social networks have become an indivisible part of our lives. To enjoy different kinds of services, people prefer to take part in multiple online social networks rather than only one. Therefore, identifying the same user across networks, formally named as social network alignment, has become a hot research topic. In this paper, we use social network structure to solve this problem. Firstly, inspired by the aligned network model (a mathematical model to formalize the real-world aligned networks), we present a novel assumption for network alignment. We suppose that the real-world aligned networks can be seen as generated from many different underlying social networks, depending on the matching between users, and the correctly aligned networks ought to own the maximum generation probability. Secondly, a Generation probability based Percolation Network Alignment method (GPNA) is presented. In GPNA, only the candidates, which can increase the generation probability, are regarded as the matched users. At last, a series of experiments are conducted to demonstrate the good performance of GPNA on both synthetic networks and real-world networks.

随着互联网行业的快速发展，在线社交网络已经成为我们生活中不可分割的一部分。为了享受不同类型的服务，人们更愿意参与多个在线社交网络，而不是只参与一个。因此，跨网络识别同一用户，正式命名为社交网络对齐，成为一个热门的研究课题。在本文中，我们使用社交网络结构来解决这个问题。首先，受对齐网络模型（一种将现实世界对齐网络形式化的数学模型）的启发，我们提出了网络对齐的新假设。我们假设现实世界的对齐网络可以被看作是由许多不同的底层社交网络生成的，这取决于用户之间的匹配，并且正确对齐的网络应该拥有最大的生成概率。第二，提出了一种基于生成概率的渗透网络对齐方法（GPNA）。在 GPNA 中，只有能够增加生成概率的候选者才被视为匹配用户。最后，进行了一系列实验以证明 GPNA 在合成网络和现实世界网络上的良好性能。