Detect Leaking Data: Safeguarding AI for Space Applications
Trust is the foundation of every mission – and when artificial intelligence enters the loop, ensuring that trust becomes essential. As part of the PINEBERRY activity (Assurance for Space Domain AI Applications), we developed a comprehensive Catalogue of Risks for AI in Space, capturing the most pressing technical and operational challenges of our time.
The “Detect Leaking Data” hackathon brings one of these real-world risks – data leakage – into focus. Participants will confront realistic scenarios where sensitive or proprietary information can leak due to the actions of malicious agents. By taking part, you will explore how such weaknesses can emerge and experiment with practical ways to detect them. In real world cases, on which our hackathon is based, time matters and you never know, which part of the system would be targeted. Stay tuned and prepare for unexpected!
What Awaits You
- Opening Session (30 min) – Learn about the background of the challenge, recent developments in AI assurance, and the nature of the data leakage phenomenon, especially for the text data.
- Hack Sprint (3 hours) – Work hands-on with provided data, test your methods, and benchmark your results in real time.
- Results & Awards – Celebrate the most innovative and effective approaches, with recognition for detection accuracy, originality, and clarity of explanation.
Why Join
- Gain firsthand experience with one of the most relevant challenges in deploying trustworthy AI for mission operations.
- Contribute insights to the projects like PINEBERRY and their effort in shaping future assurance methods for AI systems in space.
- Meet and collaborate with peers from academia, industry, and the space community in a dynamic, time-bounded competition.
Please register your interest by sending an email to pineberry@kplabs.pl or evridiki.ntagiou@esa.int with the subject: Detect Leaking Data Hackathon.
