Two-Phase Multi-Party Computation Enabled Privacy-Preserving Federated Learning

Abstract

Countries across the globe have been pushing strict regulations on the protection of personal or private data collected. The traditional centralized machine learning method, where data is collected from end-users or IoT devices, so that it can discover insights behind real-world data, may not be feasible for many data-driven industry applications in light of such regulations. A new machine learning method, coined by Google as Federated Learning (FL) enables multiple participants to train a machine learning model collectively without directly exchanging data. However, recent studies have shown that there is still a possibility to exploit the shared models to extract personal or confidential data. In this paper, we propose to adopt Multi-Party Computation (MPC) to achieve privacy-preserving model aggregation for FL. The MPC-enabled model aggregation in a peer-to-peer manner incurs high communication overhead with low scalability. To address this problem, the authors proposed to develop a two-phase mechanism by 1) electing a small committee and 2) providing MPC-enabled model aggregation service to a larger number of participants through the committee. The MPCenabled FL framework has been integrated in an IoT platform for smart manufacturing. It enables a set of companies to train high quality models collectively by leveraging their complementary data-sets on their own premises, without compromising privacy, model accuracy vis-a'-vis traditional machine learning methods and execution efficiency in terms of communication cost and execution time.

Privacy-Aware Resource Sharing in Cross-Device Federated Model Training for Collaborative Predictive Maintenance

“…In the context of PdM, the data-sets generated by two digital twins are statistically homogeneous if they contain similar failure patterns of the corresponding assets working under the same operational conditions. While existing approaches are few and limited to the application of FL [19] [20] [21] [22] without considering the computational challenges in manufacturing environment, it has been observed that these solutions are able to tackle the data privacy issue in collaborative prognosis. The collaborative prognosis mechanism proposed in [19] deploys a digital twin based FL solution that predicts the RUL of an aircraft engine.…”

“…The above discussed approaches do not address the challenges of applying conventional FL within a cross-device setting for collaborative PdM. The approach proposed in [22] is designed for cross-silo FL setting. Whereas the one proposed in [19] utilises LSTM as the global learning model which is not suitable for cross-device FL setting.…”

Secure and Provenance Enhanced Internet of Health Things Framework: A Blockchain Managed Federated Learning Approach

“…Many governments around the globe have imposed restrictions on centralized data collection processes. Hence, traditional deep learning applications that rely on a central, powerful cloud machine that accumulates a vast amount of IoHT data, and on training a model accurately are encountering regulatory restrictions [4] . In addition to regulatory restrictions, cloud-based central machine learning applications might not be suitable for training a massive amount of IoHT data [5] .…”

“…Before the start of the FL aggregation process, non-benign training results can be filtered out, and only those malicious edge nodes can be blacklisted [41] . In another study [4] , a two-phase FL process was proposed in which Phase 1 allowed voting for some trusted committee member from the federation, and in Phase 2 the actual FL takes place under the guardianship and privacy-protecting watchful eye of the committee members.…”

Applying Federated Learning in Software-Defined Networks: A Survey

“…Zhu et al [97] analyzed the relationship between FL and SMC, and Bonawitz et al [98] designed a secure communication protocol that could specify a constant number of iterations and tolerate a number of clients exiting the FL. Kanagavelu et al [99] proposed a two-phase MPC-enabled FL framework, wherein participants elected a committee to offer SMC services to reduce the overhead of SMC and maintain the scalability of FL. In [100], Sotthiwat et al suggested encrypting the key parameters of multi-dimensional gradients to take advantage of SMC encryption while limiting communication cost.…”