Model-based Anomaly Pattern Detection and Analysis in Ubiquitous and Social Interaction Networks II (MODUS II)

• Applicant: Professor Dr. Martin Atzmüller
• Subject Area: Image and Language Processing, Computer Graphics and Visualisation, Human Computer Interaction, Ubiquitous and Wearable Computing
• Term: since 2016
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 316679917

Detecting anomalies in ubiquitous and social interaction networks is an important and challenging task, in particular relating to complex (group) anomalies, as well as providing interpretable patterns. In general, anomalies can be regarded as points or patterns observed in the data that do not conform to some well-defined notion of normal behavior. However, there is usually no clear formalization of the "normal behavior". Furthermore, current research mostly targets point anomalies, i.e., only relating to individual data points; this does not include anomalies with a more complex structure, e.g., group anomalies. The main focus of the project concerns complex information and knowledge processes on complex ubiquitous and social interaction networks. Here, methods for model-based anomaly pattern detection will be investigated and developed, i.e., using explicitly formalized reference models for normal (and/or abnormal) interactions. In particular, this includes dynamic attributed interaction networks, e.g., online and offline social networks, socio-spatial sensor-based networks, as well as cyber-physical industrial network structures. With the appropriate model-based formalizations, this then allows to detect anomaly patterns more accurately, to reduce artifacts, and to increase transparency. This is enabled by respective machine learning and data analysis methods that are explainable. As a main result of the project, according machine learning and data analysis methods are developed, such that complex network data can then be comprehensively modeled and analyzed regarding model-based anomaly pattern detection, while enhancing applicability, transparency and comprehensibility of the applied approaches.

• DFG Programme: Research Grants