TITAN

Trusted AI systems through confidential computing and privacy-preserving technologies

Project Details

Duration

1 February 2024 - 31 January 2027

Total Budget

€5 Million

Status

Active

Funding Program

Horizon Europe

Project Coordinator

UNIVERSIDAD DE MURCIA

Project Description

TITAN is a 36-month project that proposes to develop secure and trustworthy confidential data processing and sharing capabilities, and demonstrate them in the EOSC ecosystem.

The sharing of sensitive data will follow FAIR data and open science principles. The project puts significant emphasis on privacy preservation and AI technological solutions in line with existing ethical, regulatory and legal EU boundaries.

The developed open-source software platform will focus mostly on the two use cases present in the project: government data and healthcare.

Project Summary

Key Objectives

Trusted AI Infrastructure: Develop secure infrastructure for AI/ML workloads using confidential computing technologies.
Privacy-Preserving Machine Learning: Enable collaborative machine learning while preserving data privacy across multiple stakeholders.
Secure Data Sharing: Implement secure multi-party computation protocols for confidential data exchange and processing.
TEE Integration: Leverage hardware-based security features for protecting AI models and training data.

Expected Outcomes

Confidential AI platform: Secure platform for training and deploying AI models in untrusted environments.
Privacy-preserving protocols: Advanced protocols for secure collaborative AI across organizations.
TEE-enabled ML: Machine learning frameworks optimized for Trusted Execution Environments.
Secure model deployment: Tools and frameworks for deploying AI models with confidentiality guarantees.

Our Involvement

Ultraviolet brings extensive expertise in confidential computing and privacy-preserving technologies to the TITAN project. Our experience with Trusted Execution Environments (TEEs), secure multi-party computation, and collaborative AI platforms positions us as a key contributor to developing trusted AI systems.

Our Responsibilities:

Confidential Computing Architecture: Designing and implementing TEE-based architectures for secure AI workloads.
Privacy-Preserving AI Frameworks: Developing frameworks for secure collaborative machine learning using SMPC and confidential computing.
Secure Model Training: Implementing secure protocols for distributed AI model training across multiple parties.
Data Protection: Ensuring data confidentiality and integrity throughout the AI lifecycle using hardware-based security.
Platform Integration: Integrating confidential computing capabilities into existing AI/ML platforms and workflows.

Learn More View on CORDIS