Privacy-preserving Cross-domain Recommendation with Federated Graph Learning This paper proposes a novel federated graph learning approach for Privacy-Preserving Cross-Domain Recommendation (PPCDR) to capture users' preferences based on distributed multi-domain data and improve recommendation performance for all domains without privacy leakage. The main idea of PPCDR is to model both global preference among multiple domains and local preference at a specific domain for a given user, which characterizes the user's shared and domain-specific tastes toward the items for interaction. Specifically, in the private update process of PPCDR, we design a graph transfer module for each domain to fuse global and local user preferences and update them based on local domain data. In the federated update process, through applying the local differential privacy technique for privacy-preserving, we collaboratively learn global user preferences based on multi-domain data and adapt these global preferences to heterogeneous domain data via personalized aggregation. In this way, PPCDR can effectively approximate the multi-domain training process that directly shares local interaction data in a privacy-preserving way. Extensive experiments on three CDR datasets demonstrate that PPCDR consistently outperforms competitive single- and cross-domain baselines and effectively protects domain privacy. The paper introduces a federated graph learning approach for Privacy-Preserving Cross-Domain Recommendation (PPCDR). The main idea of PPCDR is to model both global preference among multiple domains and local preference at a specific domain for a given user. These two kinds of preferences characterize users' shared or domain-specific tastes toward the items for interaction and are more effectively integrated through Graph Neural Networks (GNNs). For each domain, we construct a domain-specific user-item interaction graph by augmenting the links that connect global and local user nodes (corresponding to global and local user preferences) and then devise a federated graph learning method based on GNNs. To learn cross-domain knowledge for recommendation in a privacy-preserving way, each training iteration of PPCDR consists of a private update process within a local domain and a federated update across multiple domains. We design a graph transfer module for each domain to perform bi-directional message exchange and propagation, where the message-passing mechanisms of GNNs in each domain can efficiently fuse global and local user preferences, as well as capture collaborative signals inherent in local user-item interactions. Then, in the federated update process, each domain applies privacy protection technique (i.e., local differential privacy) on the learned global user preferences to enhance privacy protection and then shares them to other domains. Meanwhile, each domain receives the global user preferences from other domains and then locally updates global user preferences by a personalized aggregation strategy for domain-specific adaptation. In this way, PPCDR can effectively approximate the multi-domain training process that directly shares local interaction data in a privacy-preserving way. Besides, we propose a periodic synchronization mechanism to reduce the communication cost brought by maintaining the global preferences across domains. Our main contributions are summarized as follows: (1) ItPrivacy-preserving Cross-domain Recommendation with Federated Graph Learning This paper proposes a novel federated graph learning approach for Privacy-Preserving Cross-Domain Recommendation (PPCDR) to capture users' preferences based on distributed multi-domain data and improve recommendation performance for all domains without privacy leakage. The main idea of PPCDR is to model both global preference among multiple domains and local preference at a specific domain for a given user, which characterizes the user's shared and domain-specific tastes toward the items for interaction. Specifically, in the private update process of PPCDR, we design a graph transfer module for each domain to fuse global and local user preferences and update them based on local domain data. In the federated update process, through applying the local differential privacy technique for privacy-preserving, we collaboratively learn global user preferences based on multi-domain data and adapt these global preferences to heterogeneous domain data via personalized aggregation. In this way, PPCDR can effectively approximate the multi-domain training process that directly shares local interaction data in a privacy-preserving way. Extensive experiments on three CDR datasets demonstrate that PPCDR consistently outperforms competitive single- and cross-domain baselines and effectively protects domain privacy. The paper introduces a federated graph learning approach for Privacy-Preserving Cross-Domain Recommendation (PPCDR). The main idea of PPCDR is to model both global preference among multiple domains and local preference at a specific domain for a given user. These two kinds of preferences characterize users' shared or domain-specific tastes toward the items for interaction and are more effectively integrated through Graph Neural Networks (GNNs). For each domain, we construct a domain-specific user-item interaction graph by augmenting the links that connect global and local user nodes (corresponding to global and local user preferences) and then devise a federated graph learning method based on GNNs. To learn cross-domain knowledge for recommendation in a privacy-preserving way, each training iteration of PPCDR consists of a private update process within a local domain and a federated update across multiple domains. We design a graph transfer module for each domain to perform bi-directional message exchange and propagation, where the message-passing mechanisms of GNNs in each domain can efficiently fuse global and local user preferences, as well as capture collaborative signals inherent in local user-item interactions. Then, in the federated update process, each domain applies privacy protection technique (i.e., local differential privacy) on the learned global user preferences to enhance privacy protection and then shares them to other domains. Meanwhile, each domain receives the global user preferences from other domains and then locally updates global user preferences by a personalized aggregation strategy for domain-specific adaptation. In this way, PPCDR can effectively approximate the multi-domain training process that directly shares local interaction data in a privacy-preserving way. Besides, we propose a periodic synchronization mechanism to reduce the communication cost brought by maintaining the global preferences across domains. Our main contributions are summarized as follows: (1) It