Thursday 27 March 2025
A new protocol has been developed that aims to secure the growing field of federated learning, a type of artificial intelligence (AI) that allows multiple organisations to collaborate on training machine learning models without sharing their data.
Federated learning is becoming increasingly important as more and more companies look to leverage AI in their businesses. However, it also poses significant security risks, particularly when it comes to the sharing of sensitive data between organisations.
The new protocol, known as Post-Quantum Blockchain-based Federated Learning (PQBFL), uses a combination of post-quantum cryptography and blockchain technology to ensure the secure exchange of data between parties.
Post-quantum cryptography is a type of encryption that is designed to be resistant to attacks from quantum computers, which are expected to become powerful enough to break many current encryption algorithms in the near future. Blockchain technology, on the other hand, provides a secure and transparent way to record transactions and data exchanges between parties.
PQBFL works by allowing organisations to share their machine learning models with each other, while keeping their underlying data private. This is achieved through the use of homomorphic encryption, which allows computations to be performed on encrypted data without decrypting it first.
The protocol also includes a mechanism for ensuring that the models being shared are genuine and have not been tampered with. This is done by using digital signatures, which allow parties to verify the authenticity of the models and ensure that they have not been modified during transmission.
One of the key benefits of PQBFL is its ability to provide both confidentiality and integrity of data. Confidentiality refers to the protection of sensitive information from being accessed or viewed by unauthorised parties, while integrity refers to the prevention of tampering with or modifying data in transit.
In addition to providing strong security guarantees, PQBFL also offers several practical advantages over other federated learning protocols. For example, it does not require all parties to have a shared secret key, which can be difficult to manage and maintain. Instead, each party generates its own public-private key pair, making the protocol more scalable and easier to implement.
PQBFL has been designed with scalability in mind, and is capable of supporting large numbers of parties and models. This makes it well-suited for applications such as healthcare and finance, where multiple organisations may need to collaborate on complex AI projects.
Overall, PQBFL represents a significant step forward in the development of secure federated learning protocols.
Cite this article: “Post-Quantum Blockchain-Based Federated Learning: A Secure Protocol for AI Collaboration”, The Science Archive, 2025.
Ai, Federated Learning, Post-Quantum Cryptography, Blockchain, Homomorphic Encryption, Digital Signatures, Confidentiality, Integrity, Scalability, Quantum Computers
