PEPFL:A framework for a practical and efficient privacy-preserving federated learning

As an emerging joint learning model,federated learning is a promising way to combine model parameters of different users for training and inference without collecting users’original data.However,a practical and efficient solution has not been established in previous work due to the absence of efficient matrix computation and cryptography schemes in the privacy-preserving federated learning model,especially in partially homomorphic cryptosystems.In this paper,we propose a Practical and Efficient Privacy-preserving Federated Learning(PEPFL)framework.First,we present a lifted distributed ElGamal cryptosystem for federated learning,which can solve the multi-key problem in federated learning.Secondly,we develop a Practical Partially Single Instruction Multiple Data(PSIMD)parallelism scheme that can encode a plaintext matrix into single plaintext for encryption,improving the encryption efficiency and reducing the communication cost in partially homomorphic cryptosystem.In addition,based on the Convolutional Neural Network(CNN)and the designed cryptosystem,a novel privacy-preserving federated learning framework is designed by using Momentum Gradient Descent(MGD).Finally,we evaluate the security and performance of PEPFL.The experiment results demonstrate that the scheme is practicable,effective,and secure with low communication and computation costs.

Lattice-based certificateless encryption scheme

Certificateless public key cryptography （CL- PKC） can solve the problems of certificate management in a public key infrastructure （PKI） and of key escrows in identity-based public key cryptography （ID-PKC）. In CL- PKC, the key generation center （KGC） does not know the private keys of all users, and their public keys need not be cer- tificated by certification authority （CA）. At present, however, most certificateless encryption schemes are based on large in- teger factorization and discrete logarithms that are not secure in a quantum environment and the computation complexity is high. To solve these problems, we propose a new certificate- less encryption scheme based on lattices, more precisely, us- ing the hardness of the learning with errors （LWE） problem. Compared with schemes based on large integer factoriza- tion and discrete logarithms, the most operations are matrix- vector multiplication and inner products in our scheme, our approach has lower computation complexity. Our scheme can be proven to be indistinguishability chosen ciphertext attacks （IND-CPA） secure in the random oracle model.

Identity-based key distribution for mobile Ad Hoc networks

An identity-based cryptosystem can make a special contribution to building key distribution and management architectures in resource-constrained mobile ad hoc networks since it does not suffer from certificate management problems. In this paper, based on a light- weight cryptosystem, elliptic curve cryptography （ECC）, we propose an identity-based distributed key-distribution protocol for mobile ad hoc networks. In this protocol, using secret sharing, we build a virtual private key generator which calculates one part of a user＇s secret key and sends it to the user via public channels, while, the other part of the secret key is generated by the user. So, the secret key of the user is generated collaboratively by the virtual authority and the user. Each has half of the secret information about the secret key of the user. Thus there is no secret key distribution problem. In addition, the user＇s secret key is known only to the user itself, therefore there is no key escrow.