eduzhai > Applied Sciences > Engineering >

XOR Mixup Privacy-Preserving Data Augmentation for One-Shot Federated Learning

  • Save

... pages left unread,continue reading

Document pages: 8 pages

Abstract: User-generated data distributions are often imbalanced across devices andlabels, hampering the performance of federated learning (FL). To remedy to thisnon-independent and identically distributed (non-IID) data problem, in thiswork we develop a privacy-preserving XOR based mixup data augmentationtechnique, coined XorMixup, and thereby propose a novel one-shot FL framework,termed XorMixFL. The core idea is to collect other devices encoded datasamples that are decoded only using each device s own data samples. Thedecoding provides synthetic-but-realistic samples until inducing an IIDdataset, used for model training. Both encoding and decoding procedures followthe bit-wise XOR operations that intentionally distort raw samples, therebypreserving data privacy. Simulation results corroborate that XorMixFL achievesup to 17.6 higher accuracy than Vanilla FL under a non-IID MNIST dataset.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×