Consider two data holders, ABC and XYZ, with graph data (e.g., social networks, e-commerce, telecommunication, and bio-informatics). ABC can see that node A is linked to node B, and XYZ can see node B is linked to node C. Node B is the common neighbour of A and C but neither network can discover this fact on their own. In this paper, we provide a two party computation that ABC and XYZ can run to discover the common neighbours in the union of their graph data, however neither party has to reveal their plaintext graph to the other. Based on private set intersection, we implement our solution, provide measurements, and quantify partial leaks of privacy. We also propose a heavyweight solution that leaks zero information based on additively homomorphic encryption.
翻译:考虑两个数据持有者,ABC 和 XYZ, 包括图形数据(例如社交网络、电子商务、电信和生物信息学)。ABC可以看到节点A与节点B相连,XYZ可以看到节点B与节点C相连。 节点B是A和C的共同邻邦,但两个网络都无法单独发现这一事实。在本文中,我们提供了两个方的计算方法,即ABC和XYZ可以在其图形数据组合中发现共同邻邦,但双方都不能向对方披露其直截面图。基于私人设置的交叉点,我们执行我们的解决方案,提供测量数据,并量化部分隐私漏洞。我们还提出了一个重量性解决方案,即根据添加式同质加密法泄漏零信息。