Computational Code-Based Privacy in Coded Federated Learning

Abstract

We propose a privacy-preserving federated learning (FL) scheme that is resilient against straggling devices. An adaptive scenario is suggested where the slower devices share their data with the faster ones and do not participate in the learning process. The proposed scheme employs code-based cryptography to ensure computational privacy of the private data, i.e., no device with bounded computational power can obtain information about the other devices' data in feasible time. For a scenario with 25 devices, the proposed scheme achieves a speed-up of 4.7 and 4 for 92 and 128 bits security, respectively, for an accuracy of 95% on the MNIST dataset compared with conventional mini-batch FL.

Federated Coded Matrix Inversion

“…At this stage, the workers have knowledge of everything they need in order to recover A, before they carry out their computation tasks. By (13), the recovery is straightforward.…”

“…There are few works that leverage CC to devise secure FL schemes. Most of these works have focused on distributed regression and iterative methods, which is the primary application for FL [9]- [13]. Below, we describe and compare these approaches to our work.…”

“…Lastly, we discuss the method proposed in [13], which is based on the McEliece cryptosystem, and moderate-density paritycheck codes. This scheme considers a communication delay model which defines stragglers as the workers who respond slower than the fastest worker, and time out after a predetermined amount of time ∆.…”

“…If this is not the case, before f ι (x) are determined, the clients could perform a data exchange phase (e.g. [13]), so that N i = N j for all i = j. By this, it follows that the number of blocks does not have to be equal to the number of clients.…”