With an increasing number of satellites in orbit it is becoming vital to be able to detect and categorise behaviour in space, including unexpected actions and anomalies. The aim of this work was to develop methods for characterising satellite behaviour, including anomaly detection related to behavioural changes. During 2024, the Swedish Defence Research Agency participated in a technical research challenge focusing on the detection and classification of satellites' behavior with machine learning algorithms. A dataset containing synthetic data for satellites in geosynchronous orbits was released for the technical challenge. The purpose of the participation was to build knowledge, and to follow the state-of-the-art research within this area. Developed models could be compared to those of other participants, and the synthetic dataset could be evaluated based on real satellite data. The following report summarises the participation in the technical challenge and highlights the lessons learned from it, and it should be seen as the first step in allowing for characterisation of satellite behaviour, with the purpose of warning about potentially harmful actions in space. Based on the results and lessons learned, a new framework for characterising satellite behaviour was developed.