01/2022 – 12/2025

CausalAnomalies

"CausalAnomalies" is a project of the "Artificial Intelligence" research area of DLR's strategic research initiative on digitalisation. The aim is to develop new algorithms for recognising causally explainable anomalies and to apply these algorithms to DLR-specific problems.

The detection of attacks and misuse based on real-time log data from the EOC User Management System (UMS) is one of the use cases from the fields of space, aviation and transport.

Particular challenges here are the increased IT security and data protection requirements of the UMS, which are taken into account by means of suitable encryption-based anonymisation and pseudonymisation without impairing the detection performance of the algorithms.

CausalAnomalies
combines anomaly detection using AI-supported processes with causal inference methods in order to not only better detect anomalies, but also to be able to explain them in a comprehensible manner. The medium-term goal is the real-time online use of the resulting algorithms beyond the scope of the EOC UMS, e.g. for maintenance, monitoring and security tasks.
DownloadDownload