Boosting Service Workflow Reliability Through Enhanced Detection of Artifact Anomalies

Enterprises widely use workflow management systems (WfMS) to generate service-based workflows by orchestrating existing services. Modern WMS enables nontechnical users to create and modify workflows via graphical tools and natural language commands. Incorrect artifact operations, such as reading a destroyed artifact, may cause execution failures, data inconsistencies, or incorrect process outcomes. Existing methods for artifact anomaly detection are limited in scope and fail to detect all possible anomalies. In this paper, we first define artifact anomalies, identifying key error patterns in artifact operations. Then, we introduce the Binary Types of Arcs (BTA) Tree, a tree-like structure that captures artifact operations and control-flow of a workflow graph. Finally, we present efficient algorithms that operate on the BTA-Tree to detect anomalies more comprehensively than previous approaches. Compared with existing methods, our approach detects more anomalies with fewer false alarms while maintaining computational efficiency.