Applicability of Machine Learning Models for Detecting Anomalous File Transfer Events in B2B Application Integrations

Information systems that are based on real-time file integrations have an essential role to improve the quality of organizations’ business process management. File transfers and data integrations between discrete systems have gained great importance. However, network and security issues have emerged due to the integration of file transfer processes and the structure of files. Thus, an effective and self-learning anomaly detection approach is needed for the file integration processes to provide the persistence of integration channels, data integrity, and availability of file transfer processes. A novel anomaly detection approach that focuses on file transfers between discrete systems is proposed in this paper. The proposed anomaly detection approach is a self-learning module for the file transfer processes. For this purpose, anomalies that occur in file transfer processes are detected by applying machine learning techniques. Four clustering-based machine learning approaches are applied to detect anomalies in the file integration processes: Elliptic Envelope, Isolation Forest, Local Outlier Factor, and One-Class Support Vector Machine.

Anahtar Kelimeler: anomaly detection; machine learning; file integration; integrity

Information systems that are based on real-time file integrations have an essential role to improve the quality of organizations’ business process management. File transfers and data integrations between discrete systems have gained great importance. However, network and security issues have emerged due to the integration of file transfer processes and the structure of files. Thus, an effective and self-learning anomaly detection approach is needed for the file integration processes to provide the persistence of integration channels, data integrity, and availability of file transfer processes. A novel anomaly detection approach that focuses on file transfers between discrete systems is proposed in this paper. The proposed anomaly detection approach is a self-learning module for the file transfer processes. For this purpose, anomalies that occur in file transfer processes are detected by applying machine learning techniques. Four clustering-based machine learning approaches are applied to detect anomalies in the file integration processes: Elliptic Envelope, Isolation Forest, Local Outlier Factor, and One-Class Support Vector Machine.

Keywords: anomaly detection; machine learning; file integration; integrity