Materials
Catalogues¶
Full catalogues are accessible for registered users from ESA Member States on the Space CODEV platform:
https://gitlab.space-codev.org/assurance-ai/secure-and-explainable-ai-catalogues
Graphical introductions to catalogues are publicly available on Zenodo:
- Introduction to Secure AI in Space https://doi.org/10.5281/zenodo.14762574
- Introduction to Explainable AI in Space: https://doi.org/10.5281/zenodo.14699439


Code¶
There are four software solutions developed within the project:
- Explainability as a Service (XaaS) for anomaly detection in satellite telemetry - LIME for telemetry channels importance and CoMTE for counterfactuals generation (code: https://gitlab.space-codev.org/assurance-ai/xaas-anomaly-detection-satellite-telemetry)
- Explainability\Security as a Service (XSaaS) for satellite telemetry forecasting - Shapley values for time series to detect data poisoning (code available soon)
- Streamlit application for explainable ship detection in SAR data - six different attribution-based methods to debug ship detection models (code available soon)
- Secure and explainable interface for LLM-based documents summarization - indirect prompt injection mitigation methods and attribution-based methods to highlight most relevant input fragments (code available soon)
Publications¶
- poster at 6th Polish Conference of Artificial Intelligence PP-RAI 2025