프라이버시 보호를 위한 적대적 서명 그래프 학습
📝 Abstract
Signed graphs with positive and negative edges can model complex relationships in social networks. Leveraging on balance theory that deduces edge signs from multi-hop node pairs, signed graph learning can generate node embeddings that preserve both structural and sign information. However, training on sensitive signed graphs raises significant privacy concerns, as model parameters may leak private link information. Existing protection methods with differential privacy (DP) typically rely on edge or gradient perturbation for unsigned graph protection. Yet, they are not well-suited for signed graphs, mainly because edge perturbation tends to cascading errors in edge sign inference under balance theory, while gradient perturbation increases sensitivity due to node interdependence and gradient polarity change caused by sign flips, resulting in larger noise injection. In this paper, motivated by the robustness of adversarial learning to noisy interactions, we present ASGL, a privacypreserving adversarial signed graph learning method that preserves high utility while achieving node-level DP. We first decompose signed graphs into positive and negative subgraphs based on edge signs, and then design a gradient-perturbed adversarial module to approximate the true signed connectivity distribution. In particular, the gradient perturbation helps mitigate cascading errors, while the subgraph separation facilitates sensitivity reduction. Further, we devise a constrained breadth-first search tree strategy that fuses with balance theory to identify the edge signs between generated node pairs. This strategy also enables gradient decoupling, thereby effectively lowering gradient sensitivity. Extensive experiments on real-world datasets show that ASGL achieves favorable privacyutility trade-offs across multiple downstream tasks. Our code and data are available in https://github.com/KHBDL/ASGL-KDD26 .
💡 Analysis
Signed graphs with positive and negative edges can model complex relationships in social networks. Leveraging on balance theory that deduces edge signs from multi-hop node pairs, signed graph learning can generate node embeddings that preserve both structural and sign information. However, training on sensitive signed graphs raises significant privacy concerns, as model parameters may leak private link information. Existing protection methods with differential privacy (DP) typically rely on edge or gradient perturbation for unsigned graph protection. Yet, they are not well-suited for signed graphs, mainly because edge perturbation tends to cascading errors in edge sign inference under balance theory, while gradient perturbation increases sensitivity due to node interdependence and gradient polarity change caused by sign flips, resulting in larger noise injection. In this paper, motivated by the robustness of adversarial learning to noisy interactions, we present ASGL, a privacypreserving adversarial signed graph learning method that preserves high utility while achieving node-level DP. We first decompose signed graphs into positive and negative subgraphs based on edge signs, and then design a gradient-perturbed adversarial module to approximate the true signed connectivity distribution. In particular, the gradient perturbation helps mitigate cascading errors, while the subgraph separation facilitates sensitivity reduction. Further, we devise a constrained breadth-first search tree strategy that fuses with balance theory to identify the edge signs between generated node pairs. This strategy also enables gradient decoupling, thereby effectively lowering gradient sensitivity. Extensive experiments on real-world datasets show that ASGL achieves favorable privacyutility trade-offs across multiple downstream tasks. Our code and data are available in https://github.com/KHBDL/ASGL-KDD26 .
📄 Content
The signed graph is a common and widely adopted graph structure that can represent both positive and negative relationships using signed edges [4,6,30]. For example, in online social networks shown in Fig. 1, while user interactions reflect positive relationships (e.g., like, trust, friendship), negative relationships (e.g., dislike, distrust, complaint) also exist. Signed graphs provide more expressive power than unsigned graphs to capture such complex user interactions.
Recently, some studies [16,22,26] have explored signed graph learning methods, aiming to obtain low-dimensional vector representations of nodes that preserve key signed graph properties: neighbor proximity and structural balance. These embeddings are subsequently applied to downstream tasks such as edge sign prediction, node clustering, and node classification. Among existing signed graph learning methods, balance theory [3] has proven effective in identifying the edge signs between the source node and multi-hop neighbor nodes. It is leveraged in graph neural network (GNN)-based models to guide message passing across signed edges, ensuring that information aggregation is aligned with the node proximity [7,17,18]. Moreover, to enhance the robustness and generalization capability of deep learning models, the adversarial graph embedding model [21,33] learns the underlying connectivity distribution of signed graphs by generating high-quality node embeddings that preserve signed node proximity. Despite their ability to effectively capture signed relationships between nodes, graph learning models remain vulnerable to link stealing attacks [13,35,42], which aim to infer the existence of links between arbitrary node pairs in the training graph. For instance, in online social graphs, such attacks may reveal whether two users share a friendly or adversarial relationship, compromising user privacy and damaging personal or professional reputations.
Differential privacy (DP) [9] is a rigorous privacy framework that guarantees statistically indistinguishable outputs regardless of any individual data presence. Such guarantee is achieved through sufficient perturbation while maintaining provable privacy bounds and computational feasibility. Existing privacy-preserving graph learning methods with DP can be categorized into two types based on the perturbation mechanism: one applies edge perturbation [23] to protect the link information by modifying the graph structure, and the other adopts gradient perturbation [36,37] to obscure the relationships between nodes during model training. However, these methods are not well-suited for signed graph learning due to the following two challenges:
• Cascading error: As illustrated in Fig. 2, balance theory facilitates the inference of the edge sign between two unconnected nodes by computing the product of edge signs along a path. However, existing methods that use edge perturbation to protect link information may alter the sign of any edge along the path, thereby leading to incorrect inference of edge signs under balance theory. Such a local error can further propagate along the path, resulting in cascading errors in edge sign inference. • High sensitivity: While gradient perturbation methods without directly perturbing edges may mitigate cascading errors, they are still ill-suited for signed graph learning because the node interdependence in signed graphs leads to high gradient sensitivity. 1 Furthermore, edge change may induce sign flips that reverse gradient polarity within the loss function (see Eq. (10) for details), resulting in higher sensitivity compared to unsigned graphs. This increased sensitivity requires larger noise for privacy protection, thereby reducing the data utility.
To address these challenges, we turn to an adversarial learningbased approach for private signed graph learning. The core motivation is that this adversarial method generates node embeddings by approximating the true connectivity distribution, making it naturally robust to noisy interactions during optimization. As a result, we propose ASGL, a differentially private adversarial signed graph learning method that achieves high utility while maintaining nodelevel differential privacy. Within ASGL, the signed graph is first decomposed into positive and negative subgraphs based on edge 1 The presence or absence of a node affects gradient updates of itself and its neighbors.
signs. These subgraphs are then processed through an adversarial learning module within shared model parameters, enabling both positive and negative node pairs to be mapped into a unified embedding space while effectively preserving signed proximity. Based on this, we develop the adversarial learning module with differentially private stochastic gradient descent (DPSGD), which generates private node embeddings that closely approximate the true signed connectivity distribution. In particular, the gradient perturbation helps mitigate cascading errors, while the subg
This content is AI-processed based on ArXiv data.