EAGER: Private Blockchain-Enabled Federated Learning Framework for Distributed Manufacturing Networks

近年来，全球制造网络经历了包括新冠肺炎疫情在内的各种冲击和干扰。因此，提高网络弹性、透明度和网络安全已成为国家的优先事项。人工智能和机器学习等智能制造技术有望实现这些目标，但在制造网络层面却难以实现。由于缺乏资源和激励措施，中小型制造商尤其难以采用这些数据驱动的增值技术。因此，他们不能参与许多高价值的制造网络，这些网络通常需要某些技术和数据共享。这个探索性研究（EAGER）项目的早期概念资助支持旨在通过区块链支持的框架来解决这一挑战的研究，该框架利用安全和私有的联邦学习来满足国防制造网络的独特要求。该框架增强了关键物资的可用性和完整性，并加强了国防工业基础并使其多样化。该项目的安全和保护隐私的数据共享和协作机制可以应用于制造业以外的各个领域，如医疗保健、金融和供应链，使个人和组织能够安全共享数据并有效协作。其成果有可能改变产业，推动经济增长，促进创新，提高社会福祉。该项目的研究问题源于制造网络无法安全有效地交换数据和利用网络级联邦学习。该项目旨在通过提供安全的私有区块链平台，实现分散、安全和透明的通信渠道，提高分布式和动态制造网络的弹性，特别是包括中小型制造商。这可以通过联邦学习实现制造网络级学习，同时尊重数据所有权并确保保留竞争或受控（原始）数据和机器学习模型。为了实现这些目标，该项目通过集成私有区块链来利用联邦学习来管理元数据、访问控制和模型更新。与现有方法不同，该框架侧重于制造网络的具体挑战和要求。这意味着确保机密数据在各个节点的完全控制下保持在本地，同时利用区块链进行联邦学习过程的有效协调，并减少资源约束的较小网络参与者的开销成本。该项目通过在制造网络中异构数据存在的情况下进行模型聚合和协调的有效算法，推进了联邦学习和区块链技术的最新发展。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

In recent years, global manufacturing networks experienced a variety of shocks and disturbances including COVID-19. Thus, improving network resiliency, transparency, and cybersecurity have emerged as a national priority. Smart Manufacturing technologies such as Artificial Intelligence and Machine Learning show promise in achieving these objectives, yet struggle to materialize at the manufacturing network level. Particularly small and medium-sized manufacturers struggle in their adoption of these data-driven, value added technologies due to a lack of resources and incentives. Consequently, they cannot participate in many high-value manufacturing networks that often require certain technologies and data sharing. This EArly-concept Grant for Exploratory Research (EAGER) project supports research that intends to address this challenge through a Blockchain-enabled framework that leverages secure and private Federated Learning which meets the unique requirements of defense manufacturing networks. This framework enhances the availability and integrity of critical supplies, as well as strengthens and diversifies the defense industrial base. The project’s secure and privacy-preserving data sharing and collaboration mechanisms can be applied in various domains beyond manufacturing, such as healthcare, finance, and supply chain, empowering individuals and organizations to share data securely and collaborate effectively. The results have potential to transform industry, drive economic growth, foster innovation, and enhance societal well-being. The project’s research problem stems from manufacturing networks’ inability to securely and efficiently exchange data and leverage network level Federated Learning. The project aims to increase the resiliency of distributed and dynamic manufacturing networks, specifically including small and medium-sized manufacturers, by providing access to a secure private Blockchain platform that enables decentralized, secure, and transparent communication channels. This enables manufacturing network level learning through Federated Learning while respecting data ownership and ensuring retention of competitive or controlled (raw) data and machine learning models. To achieve these goals, the project utilizes Federated Learning by integrating a private Blockchain to manage metadata, access controls, and model updates. Unlike existing approaches, the framework focuses on specific challenges and requirements of manufacturing networks. This means ensuring confidential data remains local under full control of the individual nodes while leveraging Blockchain for efficient coordination of the Federated Learning process as well as reducing overhead cost for smaller network participants that are resource constraint. The project advances the state-of-the-art in Federated Learning and Blockchain technology through efficient algorithms for model aggregation and coordination in the presence of heterogeneous data for manufacturing networks.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.