Dimensions overall score 7.0

GitHub Code Pulse

No public code linked for this paper yet.

• Existing Hybrid Homomorphic Encryption Federated Learning (HHE-FL) systems rely on a single homomorphic key pair shared across all clients, leading to a weak threat model where a single misbehaving client can expose private updates.limitation 95%
• Masking client keys before homomorphic encryption and unblinding them homomorphically by the server is a mechanism to protect against key misuse by other clients in HHE-FL.method 95%
• RSA encapsulation, where homomorphically encrypted keys are wrapped under the server's RSA public key, is another mechanism to prevent key misuse by other clients in HHE-FL.method 95%
• The proposed HHE-FL approaches (masking and RSA encapsulation) preserve model accuracy while adding minimal overhead.result 90%
• The proposed HHE-FL approaches extend security to adversarial settings with malicious participants.technical 90%
• Both masking and RSA encapsulation mechanisms were implemented on top of the Flower framework using the PASTA/BFV HHE scheme.method 95%
• The evaluation of the proposed HHE-FL mechanisms was conducted on the MNIST dataset with 12 clients.method 95%
• Existing Hybrid Homomorphic Encryption (HHE) Federated Learning (FL) systems rely on a single homomorphic key pair shared across all clients, creating a weak threat model vulnerable to key misuse by other clients.limitation 95%
• Masking client keys before homomorphic encryption and unblinding them homomorphically by the server prevents key misuse by other clients in HHE-FL.method 95%
• RSA encapsulation, where homomorphically encrypted keys are wrapped under the server's RSA public key, prevents key misuse by other clients in HHE-FL.method 95%
• Both masking and RSA encapsulation mechanisms preserve model accuracy in HHE-FL.result 90%
• The masking mechanism incurs negligible overhead in HHE-FL.result 90%

Competitive landscape

Competitor map is still being generated for this paper. Enable generation or check back soon.

Startup potential card