Intel® Tiber™ Secure Federated AI
Protect sensitive data and intellectual property while improving model accuracy.
Now in Beta – A Turnkey Federated Learning Service for Training AI Models on Private Data
Builders require diverse, real-world datasets to create robust and generalizable AI models, but privacy regulations make it difficult to obtain datasets based on private and sensitive data. Federated learning offers a solution, but the architecture can be difficult to scale, manage, operate, and deploy.
These challenges are why Intel developed Intel Tiber Secure Federated AI, a turnkey service designed to securely train AI models on private data using federated learning.
Get Started
Protect sensitive data and intellectual property while improving model accuracy.
Intel® Tiber™ Trust Services
Learn more about the full portfolio.
Contact a Consultant
Get the conversation started with an Intel representative to discuss the best deployment for your unique needs.
Product Benefits
Intel Tiber Secure Federated AI uses hardware-based security, cryptography, and algorithmic techniques to help ensure high levels of security for both models and data. By providing a turnkey implementation of OpenFL, the service offers several key advantages for our customers.
Enhanced Security and Privacy
Implements zero-trust security measures to protect sensitive data and model intellectual property.
Improved Model Accuracy
Train your AI products on a larger and more diverse dataset to enhance quality and generalizability.
Operational Efficiency
Control costs by reducing data duplication and redaction efforts.
Regulatory Compliance
Enforce access control around who is using your data and how it is being used.
Building On the Foundation of OpenFL
Intel Tiber Secure Federated AI is built on OpenFL, an open source federated learning framework.
Instead of sending data to a central server, federated training allows model training locally on each device, with only the model updates being shared and aggregated to improve the overall model. This helps preserve data privacy and security, comply with data sovereignty requirements, and protect intellectual property.
OpenFL has been widely used across multiple industries and is the only federated learning framework approved for use on the International Space Station.
Use Cases
Collaborative Medical Research
Healthcare providers, hospitals, clinics, and health research companies have used AI and Machine Learning models to analyze millions of patient datasets to help better understand, predict, and prevent various types of diseases and illnesses.
Intel Tiber Secure Federated AI can be used to train AI/ML models, allowing multiple parties to contribute to the model by bringing the algorithm to the data. This helps improve clinical results, since the AI models are more robust and generalizable, while hospitals and clinics maintain control of sensitive patient data.
Early Drug Discovery
The discovery and development of novel therapeutics is a resource-intensive process that requires specialized domain expertise. Generative machine learning models have emerged as powerful tools for drug discovery, but their performance and generalizability are heavily dependent on data that is often siloed across different research institutions and companies.
Combining this data could capture a more comprehensive and representative distribution, leading to a more robust model. However, this is not feasible due to privacy and other legal concerns, competitive pressure, and technical constraints.
Intel Tiber Secure Federated AI can be used to train these models without combining datasets. This enables companies to securely collaborate on model training while managing concerns about data privacy.
Fraud Detection
AI and machine learning are increasingly being used to detect fraud in real time, but many small and medium-sized banks do not have the volume of transaction data needed to train a robust detection model. Multiple banks could pool their fraud data but cannot do so due to regulatory concerns.
Intel Tiber Secure Federated AI can be used to securely train fraud detection models across multiple banks without moving data. This can help reduce losses with more accurate fraud detection.
FAQs
Frequently Asked Questions
Intel® Tiber™ Secure Federated AI is a turnkey service designed to securely train AI models on private data using federated learning. It helps ensure that data always stays in the data owner's custody no matter where it is stored – on-premises, public cloud, or private cloud. The service uses hardware-based security, cryptographic methods, and algorithmic techniques to help ensure high levels of privacy and security for both models and data.
Federated learning (FL) is a machine learning technique that enables AI models to be trained across multiple decentralized devices or servers holding local data samples, without moving them. Instead of sending data to a central server, federated training allows the model to be trained locally on each device, with only the model updates being shared and aggregated to improve the overall model. This method increasingly preserves data privacy and security, helps ensure compliance with data sovereignty laws, and offers enhanced protections of intellectual property.
Intel Tiber Secure Federated AI is built on OpenFL, an open source federated learning framework developed by Intel as part of the Linux Foundation LF AI and Data project. OpenFL has been widely used across industries such as insurance, pharmaceuticals, and healthcare, and is the only federated learning framework approved for use on the International Space Station.
By providing a turnkey implementation of OpenFL, Intel Tiber Secure Federated AI offers two key advantages for our customers:
Simplified configuration: Provides a user-friendly setup process that helps reduce the complexity and time required to establish federated learning environments.
Enhanced security features: Implements zero-trust security measures designed to protect sensitive data and model intellectual property.
Model builders require diverse, real-world datasets to create robust and generalizable AI models. Intel Tiber Secure Federated AI is designed to improve model development with secure, privacy-preserving techniques that help organizations collaboratively train models on distributed data.
Intel Tiber Secure Federated AI is designed to enable data collaboration by allowing institutions to train AI models using decentralized data while keeping it secure and private. The service uses hardware-based security (including confidential computing and hardware and workload attestation), cryptographic methods, and algorithmic techniques designed for high levels of privacy and security for both models and data.
Your data is stored locally at each data site or institution in a federated AI system, where data owners maintain full custody of their sensitive datasets without needing to centralize them. This decentralized approach helps ensure your data remains secure and private while allowing for collaborative model training and evaluation.
Please contact your Intel representative for more information on our product and beta program.