【INFOCOM2020】De-anonymization of Social Networks: the Power of Collectiveness

De-anonymization of Social Networks: the Power of Collectiveness

百度翻译 -> 社交网络的去匿名化：集体性的力量

1.社交网络去匿名化的定义:

Such process of unveiling users’ identities by leveraging information from other domains is defined as social network de-anonymization, which is initiated by Narayanan and Shmatikov [3].

这种通过使用来自其他领域的信息公开用户身份的过程被定义为社交网络去匿名化，这是由Narayanan和Shmatikov提出[3]。

2.社交网络去匿名化的挑战:

On the one hand, if we depend merely on the adjacency relationships [8] in trade for efficiency, the local topologies of nodes may not bring desirable matching accuracy in solving this problem, as existing researches [5], [6], [9] on different network models usually set demands of a large mean degree . On the other hand, to explore richer structural commonalities between
two networks, the similarity matrix or Kronecker product [10] is often referred, which, however, may bring a high computational complexity of . Therefore, the question naturally arises: is it possible to solve the de-anonymization problem both effectively and efficiently?

一方面，如果我们仅仅使用邻接关系去保证去匿名化的效率，局部节点的拓扑不会带来可取的匹配精度，已有的研究表示计算杂度为平均每个节点 ；另一方面，如果使用两个网络之间的富结构信息的相似度，会产生极大的计算复杂度为 。因此，问题为:能否既有效率，有足够精确地解决去匿名化的问题？

3.社交网络的一个有趣的现象

To answer this question, we note that in social networks,there is an interesting phenomenon named friendship paradox[11], where, on average, the number of friends of our random friend is always greater than or equal to the number of ourselves’ friends. Put differently, the neighbors of a user may expose more information than the user himself/herself.

“朋友悖论”

Particularly, in the common paradigm mentioned earlier, we collect the multi-hop neighborhood relations between different node pairs in both the sanitized and anonymized networks, and aim to match the users in the two networks by minimizing the total number of the mismatched relations within l hops. This number is called collective adjacency disagreement (CAD) at level l.

使用不同层级跳数的邻接节点的相似度进行匹配。