Privacy-Enhanced and Verification-Traceable Aggregation for Federated Learning

Yanli Ren, Yerong Li, Guorui Feng, Xinpeng Zhang

Abstract:

Federated learning (FL) is a distributed machine
learning framework, which allows multiple users to collaboratively
train and obtain a global model with high accuracy. Currently,
FL are paid more attention by researchers and a growing
number of protocols are proposed. This paper first analyzes the
security vulnerabilities of the VerifyNet and VeriFL protocol, and
we propose a new FL protocol which can enhance the privacy
security including privacy of local data and aggregation results.
The proposed protocol can also achieve traceable verification,
which means the users can not only verify the aggregation results,
but also be able to identify the wrong epoch if the results are
wrong. We use additive homomorphic encryption and doublemasking
to simultaneously protect the user’s local model and the
aggregated global model. Also, linear homomorphic hash and
digital signature are used to verify the result returned by the
server and identify the wrong epoch. The experimental results
show that our protocol has better security performance under
the same efficiency level.