Advantages:
- Protects user data during federated learning, ensuring individual privacy without compromising analysis.
- Minimizes data transfer, improves federated learning convergence speed and reducing resource consumption.
- Requires just one trustworthy assisting node, making implementation and scalability more straightforward.
Summary:
This technology, known as FESA, simplifies secure aggregation protocols for privacy-preserving federated learning. FESA allows servers to calculate the sum of user gradients while keeping each user's data private. Unlike traditional methods, FESA significantly reduces the amount of data exchanged between users and servers, making it faster and more efficient. The unique feature of FESA is its use of assisting nodes aiding the central server in unmasking the users’ gradients. FESA simply extends the trust level to the set of assisting nodes and only require a single assisting nodes to be honest of the security of the system to hold. In a nutshell, FESA enables a faster convergence in federated learning models and make them safer and more efficient for all the involved entities (users, servers and assisting nodes).
Overview of FESA system
Desired Partnerships:
- License
- Sponsored Research
- Co-Development