Robust blockchain-based federated learning

Akram, Aftab; Gritti, Clémentine; Halip, Mohd Hazali Mohamed; Kamarudin, Nur Diyana; Mansor, Marini; Rahayu, Syarifah Bahiyah; Önen, Melek
ICISSP 2025, 11th International Conference on Information Systems Security and Privacy, 20-22 February 2025, Porto, Portugal

In Federated Learning (FL), clients collaboratively train a global model by updating it locally. Secure Aggregation (SA) techniques ensure that individual client updates remain protected, allowing only the global model to be revealed while keeping the individual updates private. These updates are usually protected through expensive cryptographic techniques such as homomorphic encryption or multi-party computation. We propose a new solution that leverages blockchain technology, specifically the Secret Network (SN), to provide privacypreserving aggregation with aggregate integrity through Smart Contracts in Trusted Execution Environments (TEEs). Moreover, FL systems face the risk of Byzantine clients submitting poisoned updates, which can degrade the model performance. To counter this, we integrate three state-of-the-art robust aggregation techniques within the Smart Contract, namely Krum, Trim Mean and Median. Furthermore, we have evaluated the performance of our framework which remains efficient in terms of computation and communication costs. We have also exhibited similar accuracy results compared to state-of-the art scheme named SABLE.

Type:
Conference
City:
Porto
Date:
2025-02-20
Department:
Digital Security
Eurecom Ref:
8026
Copyright:
© Springer. Personal use of this material is permitted. The definitive version of this paper was published in ICISSP 2025, 11th International Conference on Information Systems Security and Privacy, 20-22 February 2025, Porto, Portugal and is available at :

PERMALINK : https://www.eurecom.fr/publication/8026