Dual-Encoder Graph Anomaly Detection via Structural–Attribute Consistency Learning

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

Static graph anomaly detection is important in applications such as cybersecurity, fraud detection, and social network analysis. However, detecting anomalies in attributed graphs remains challenging because of complex graph structures and limited labeled data. Existing methods often focus on either structural information or node attributes separately, which may reduce detection performance. This paper proposes a dual-encoder framework for anomaly detection in static graphs. The proposed method combines a GraphSAGE-based structural encoder with node-level attention and an MLP-based attribute encoder to learn complementary node representations. In addition, a hybrid anomaly scoring mechanism is introduced to capture both structural–attribute inconsistency and neighbor-aware deviations. Experiments on benchmark datasets including Cora, Citeseer, and PubMed demonstrate that the proposed method achieves competitive performance compared to existing approaches.