A Hybrid Approach to Privacy-Preserving Federated Learning
By Stacey Truex, Nathalie Baracaldo, Ali Anwar, Thomas Steinke, Heiko Ludwig, Rui Zhang. 2018
1,112 citations recorded by the source corpus.
Federated learning facilitates the collaborative training of models without the sharing of raw data. However, recent attacks demonstrate that simply maintaining data locality during training processes does not provide sufficient privacy guarantees. Rather, we need a federated learning system capable of preventing inference over both the messages exchanged during training and the final trained model while ensuring the resulting model also has acceptable predictive accuracy. Existing federated learning approaches either use secure multiparty computation (SMC) which is vulnerable to inference or differential privacy which can lead to low accuracy given a large number of parties with relatively small amounts of data each. In this paper, we present an alternative approach that utilizes both differential privacy and SMC to balance these trade-offs. Combining differential privacy with secure multiparty computation enables us to reduce the growth of noise injection as the number of parties increases without sacrificing privacy while maintaining a pre-defined rate of trust. Our system is therefore a scalable approach that protects against inference threats and produces models with high accuracy. Additionally, our system can be used to train a variety of machine learning models, which we validate with experimental results on 3 different machine learning algorithms. Our experiments demonstrate that our approach out-performs state of the art solutions.
Subfields: Artificial Intelligence, Federated & Distributed Learning
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Artificial Intelligence
- Qiang Yang
- Percy Liang
- Philip S. Yu
- Quoc V. Le
- Yang Liu
- Mohit Bansal
- Ruslan Salakhutdinov
- Lichao Sun
In Federated & Distributed Learning
- Qiang Yang
- Murali Annavaram
- Chaoyang He
- Bingsheng He
- Qinbin Li
- Yang Liu
- S. Avestimehr
- H. Vincent Poor
Citations in the corpus
- Federated Learning With Differential Privacy: Algorithms and Performance Analysis (2019, background): “The work in [35] presented an alternative approach that utilizes both DP and secure multiparty computation (SMC) to prevent differential attacks.”
- The Distributed Discrete Gaussian Mechanism for Federated Learning with Secure Aggregation (2021, background): “For instance, the majority of existing works ignore the finite precision and modular summation arithmetic associated with secure aggregation (Goryczka et al., 2013; Truex et al., 2019; Valovich & Alda, 2017).”
- FedML: A Research Library and Benchmark for Federated Machine Learning (2020, methodology): “…robust [13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 22], and proposing methods such as differential privacy (DP) and secure multiparty computation (SMPC) to protect privacy [25, 26, 27, 28, 29, 30, 31, 32, 33] (see Table 8 in the appendix for a taxonomy of the federated learning literature).”
- IBM Federated Learning: an Enterprise Framework White Paper V0.1 (2020, methodology): “Methods of creating differentially private noise or secure multi-party computation are used to meet these privacy requirements [8, 13, 31, 33].”
- Differentially Private Meta-Learning (2019, methodology): “…as FedAvg (McMahan et al., 2017), have inspired many works also concerning DP in a multi-user setup (Agarwal et al., 2018; Bhowmick et al., 2019; Geyer et al., 2018; McMahan et al., 2018; Truex et al., 2019), we are the first to consider task-global DP as a useful variation on standard DP settings.”
- Federated Model Distillation with Noise-Free Differential Privacy (2020, background): “2018), or local differential privacy (LDP), in which each user randomizes its gradients locally before sending it to an untrusted aggregator (Truex et al. 2019; Sun et al. 2020).”
- A Survey on Federated Learning: The Journey From Centralized to Distributed On-Site Learning and Beyond (2021, methodology): “[111] proposed to implement both DP and SMC in an hybrid approach.”
- Federated Machine Learning: Survey, Multi-Level Classification, Desirable Criteria and Future Directions in Communication and Networking Systems (2021, background): “The authors of [168] argue that the approaches that use differential privacy only end up with low accuracy in the presence of a large number”
- A Survey on Federated Learning Systems: Vision, Hype and Reality for Data Privacy and Protection (2019, background): “[168] combine both secure multiparty computation and differential privacy for privacy-preserving FL.”
- Towards Taming the Resource and Data Heterogeneity in Federated Learning (2019, background): “To address this requirement, very recently Federated Learning (FL) has emerged as an alternative way to do collaborative model train models without sharing the training data [12] [17] [18].”
- Towards Federated Graph Learning for Collaborative Financial Crimes Detection (2019, methodology): “…6 UK banks’ local training processes and one process was used as a proxy for the role of the aggregator using our federated learning framework [Truex et al., 2018] We have tried to train several types of machine learning models, for example, (cid:96) -1 regularized logistic regression,…”
- D2P-Fed: Differentially Private Federated Learning With Efficient Communication (2020, methodology): “To improve model utility, secure multiparty computation (SMC) has been used in tandem with DP to reduce noise (Jayaraman et al., 2018; Truex et al., 2019).”
- Lightweight Crypto-Assisted Distributed Differential Privacy for Privacy-Preserving Distributed Learning (2020, background): “More recently, Truex et al. [11] recommended a hybrid approach to privacy-preserving federated learning.”
- LDP-Fed: federated learning with local differential privacy (2020, methodology): “Existing solutions to protect FL systems from such privacy attacks require trusted aggregators [17] or heavy cryptographic techniques [6, 23] which do not allow individual participants to define different local privacy guarantees, are insufficient for meaning-fully protecting each high dimensional…”
- Federated Learning for Internet of Things: Recent Advances, Taxonomy, and Open Challenges (2020, methodology): “Therefore, to resolve these issues, one can use a hybrid approach that combines both local differential privacy via the addition of minimum possible noise and secure multiparty computation-based scheme via low-overhead encryption [106].”
- Federated Learning for Healthcare: Systematic Review and Architecture Proposal (2022, background): “[76] explore multiple machine learning algorithms to classify EHR data from nurseries.”
- Federated Learning for Healthcare Informatics (2019, background): “Thus, some researchers combine DP with SMC to reduce the growth of noise injection as the number of parties increases without sacrificing privacy while preserving provable privacy guarantees, protecting against extraction attacks and collusion threats [18, 100].”