| Federated learning(FL),as a new distributed machine learning method,is widely applied in various industries.FL allows multiple users to collaborate in training a global model by only uploading their locally trained local model parameters to the aggregator,without their own dataset leaving the local.Therefore,FL is suitable for some data security-sensitive fields such as financial institutions,medical institutions.Although FL can protect user privacy to some extent,there are still some security issues,including privacy of user data and accuracy of model aggregation.To address these two issues,this paper defines strong security federated learning,namely privacy-preserving federated learning(PPFL)and verifiable federated learning(VFL).The main work is as follows:1.Proposes a privacy-preserving federated learning scheme.Existing PPFL schemes are based on differential privacy(DP),multi-party computation(MPC),homomorphic encryption(HE)and functional encryption(FE).Compared with DP,MPC and HE,FE-based schemes can not only avoid the loss of model accuracy caused by DP,but also avoid the additional communication overhead caused by MPC and HE,which is more advantageous in PPFL scenarios,thus become the focus of this work.Moreover,all existing PPFL schemes via FE employ a multi-user extension of FE for a specific function,i.e.,multi-input FE(MIFE).In this paper,we point out that existing FE-based PPFL schemes have faced with several security issues due to the misuse of MIFE.After reconsidering the security requirements of PPFL,we propose new goals of designing PPFL using FE.To achieve our goals,we propose a new FE called dual-mode decentralized multi-client FE(2DMCFE)and give a concreate construction for 2DMCFE.With 2DMCFE,we propose a new framework of PPFL where we establish a fresh 2DMCFE instance for each subset of users.Security proof shows the strong security of our framework under the semi-honest security setting.Furthermore,experiments conducted on real dataset demonstrate that our framework achieves comparable model accuracy and training efficiency to the basic FE-based scheme while providing stronger security guarantee.2.Proposes a verifiable federated learning scheme.Existing VFL schemes are based on homomorphic hash(HH),lagrange interpolation(LI)and double aggregation(DA).Compared with HH and LI,DA-based schemes can not only avoid the security issues caused by HH,but also avoid the additional communication overhead caused by LI,which is more advantageous in VFL scenarios,thus become the focus of this work.In addition,current DA-based schemes generally use MPC as the underlying algorithm.In this paper,we combine DA with decentralized multi-client functional encryption(DMCFE)to propose a secure and verifiable federated learning framework.Due to the strong security of DMCFE,we can solve the brute force issues that exist in most VFL scenarios.In addition,we take advantage of the parameters in DMCFE,i.e.labels that identify each training rounds,to achieve verifiability without introducing extra communication between users.Experiments show that our proposed framework not only provides stronger security,but also achieves higher training efficiency than the existing scheme. |
