Privacy-preserving video analytics: Hungarian development at the international forefront
We are honored that our team won the Best Paper Award at the international CYBER 2025 conference (Lisbon, Portugal) with our research paper „Privacy-preserving Video Processing Pipeline”. The goal of the project was to build a privacy-preserving, scalable video analytics system that can recognize objects, faces and movements across multiple camera feeds in real time, while complying with the requirements of the GDPR and the EU AI Act.
What makes this solution special?
Most video surveillance systems only record, but do not analyze. Our approach brings real intelligence behind the cameras: processing happens on-premise, so sensitive data never goes to the cloud, and the system can run real-time analysis even with minimal bandwidth.
The solution is built on open-source technologies (Apache Airflow, Superset, YOLOX, FaceNet and others) and offers features such as people counting, route and heatmap analysis, and geofencing. Results appear on interactive, easy-to-read dashboards that support data-driven decision-making in retail, logistics and smart infrastructure.
What comes next
Our research showed that video analytics and data protection can go hand in hand. The next step is to integrate federated learning, anomaly detection and active machine learning techniques, which will further improve the accuracy and security of the system.
We thank the CYBER 2025 organizers and the professional community for the recognition. This award is a huge motivation for our further development work.
For more details and the technological background, visit the VeriDome page.