This is a page that curates AI-related papers published worldwide. All content here is summarized using Google Gemini and operated on a non-profit basis. Copyright for each paper belongs to the authors and their institutions; please make sure to credit the source when sharing.
Setup Once, Secure Always: A Single-Setup Secure Federated Learning Aggregation Protocol with Forward and Backward Secrecy for Dynamic Users
Created by
Haebom
Author
Nazatul Haque Sultan, Yan Bo, Yansong Gao, Seyit Camtepe, Arash Mahboubi, Hang Thanh Bui, Aufeef Chauhan, Hamed Aboutorab, Michael Bewong, Dineshkumar Singh, Praveen Gauravaram, Rafiqul Islam, Sharif Abuadbba
Outline
This paper presents a novel secure aggregation protocol that enhances privacy in federated learning (FL). To address the limitations of existing single-configuration protocols, such as lack of dynamic user participation and forward and backward security, we leverage lightweight symmetric homomorphic encryption and key negation techniques to efficiently mask updates without user-to-user communication. Furthermore, we introduce a lightweight verification mechanism using message authentication codes (MACs) to protect against model mismatch attacks. We provide formal security proofs under semi-honest and malicious adversarial models, and present a prototype implementation and experimental results to demonstrate the protocol's practicality. Experimental results demonstrate that our protocol reduces user-side computation by up to 99% compared to existing state-of-the-art protocols (e-SeaFL) while maintaining competitive model accuracy.
Takeaways, Limitations
•
Takeaways:
◦
We present an efficient federated learning secure aggregation protocol that supports dynamic user engagement and forward/backward security with a single setup.
◦
Efficient update masking without user-to-user communication is achieved by utilizing lightweight symmetric homomorphic encryption and key negation techniques.
◦
Defending against model mismatch attacks through a lightweight verification mechanism based on message authentication codes (MACs).
◦
Reduces user-side computation by up to 99% compared to existing state-of-the-art protocols while maintaining competitive model accuracy.
◦
Suitable for real-world federated learning deployments on resource-constrained devices.
•
Limitations:
◦
Further research is needed to determine the long-term stability and scalability of the proposed protocol in real-world application environments.
◦
The need for enhanced security analysis against more diverse and complex adversarial attack scenarios.
◦
Further optimization research is needed to improve the performance of the protocol.
