Private set intersection (PSI) facilitates the computation of intersection between the private sets of two parties, ensuring that no additional information beyond the intersection itself is revealed. However, most state-of-the-art are limited to static PSI, leaving updatable PSI untouched. Existing PSI protocols will cost huge computational resources to compute intersection on updated sets. More seriously, none of the existing updatable PSI approaches can achieve both secure addition and deletion operations in once update. To address these challenges, we propose Forward Private Updatable PSI (FUPSI) for two-party setting. FUPSI is designed to support addition and deletion simultaneously, while ensuring forward privacy against semi-honest adversaries. In this work, we analyze the infeasibility of secure synchronous addition and deletion in the existing updatable PSI approaches, by presenting a practical attack which would lead to privacy leakages while deletion function is performed. Then, to resist this attack against semi-honest adversaries, we demonstrate how FUPSI can protect the forward privacy of user sets, by utilizing a variant of keyword Private Information Retrieval (PIR) to hide sensitive intermediate parameters. Specifically in FUPSI, two parties execute keyword PIR to retrieve a flag indicating that the current element is added or deleted so as to determine whether it is in the participants’ datasets. Finally, we provide the formal security proof for our proposed FUPSI, and extensive experimental results demonstrate efficiency and the practicality of our proposal. For instance, the communication complexity of our proposal is only logarithmically related to the size of update sets and the computational overhead is mainly composed of logarithmical times PIR calculations. Owing to the variant of keyword PIR, our work also incurs minimal communication overhead even for enormous datasets, which performs well in updatable settings and slow networks.

中文翻译：

---

可更新的私有集与前向隐私的交集


私有集交集（PSI）有助于计算两方私有集之间的交集，确保除了交集本身之外不会泄露任何其他信息。然而，大多数最先进的技术仅限于静态 PSI，而可更新的 PSI 则保持不变。现有的 PSI 协议将花费大量的计算资源来计算更新集上的交集。更严重的是，现有的可更新PSI方法都无法在一次更新中同时实现安全的添加和删除操作。为了应对这些挑战，我们提出了用于两方设置的前向私有可更新 PSI (FUPSI)。 FUPSI 旨在同时支持添加和删除，同时确保针对半诚实对手的前向隐私。在这项工作中，我们通过提出一种在执行删除功能时会导致隐私泄露的实际攻击，分析了现有可更新 PSI 方法中安全同步添加和删除的不可行性。然后，为了抵御这种针对半诚实对手的攻击，我们演示了 FUPSI 如何通过利用关键字私有信息检索（PIR）的变体来隐藏敏感的中间参数，从而保护用户集的前向隐私。具体来说，在FUPSI中，两方执行关键字PIR来检索指示当前元素被添加或删除的标志，以确定其是否在参与者的数据集中。最后，我们为我们提出的 FUPSI 提供了正式的安全证明，大量的实验结果证明了我们建议的效率和实用性。 例如，我们建议的通信复杂度仅与更新集的大小成对数关系，并且计算开销主要由对数次 PIR 计算组成。由于关键字 PIR 的变体，即使对于巨大的数据集，我们的工作也会产生最小的通信开销，这在可更新设置和慢速网络中表现良好。