A distributed system log anomaly detection framework based on graph-chain architecture
Mingyu Hu, Song Chen, Tao Xie · Advances in Engineering Innovation · 2025
Enterprise alliances and government agencies typically deploy distributed systems across subsidiaries and subordinate departments, where each system node collaborates via a network to present a unified interface to users. Addressing the challenges of log data dispersion, heterogeneity, and lack of credibility in distributed systems, this paper proposes a log anomaly detection framework based on a graph-chain architecture. The framework leverages the sequence analysis capabilities of a distilled Transformer model to detect anomalies in system logs at each node. Finally, by integrating blockchain smart contracts, it ensures tamper resistance and traceability. Experimental results demonstrate that the proposed framework achieves an anomaly detection accuracy of 99.6%, surpassing traditional methods.