Recently, the IEEE (Institute of Electrical and Electronics Engineers) officially released IEEE P3127 Guide for an Architectural Framework for Blockchain-based Federated Machine Learning - the industry's first international standard integrating federated learning with blockchain technology. This standard was jointly developed by AsiaInfo Technologies as the lead organization, in collaboration with more than ten renowned enterprises and universities, such as the Institute for AI Industry Research (AIR) at Tsinghua University, China Telecom, China Mobile, China Unicom, Beijing University of Posts and Telecommunications, and the Chinese University of Hong Kong.
Image: Release of IEEE P3127 International Standard
The IEEE P3127 standard provides a unified architectural framework guide for blockchain-based federated learning, significantly enhancing the security, traceability, and privacy protection capabilities of multi-party collaborative modeling. This standard effectively addresses the challenges of "data silos" and "trust deficits" in cross-institutional collaboration, offering secure and reliable data-sharing and model optimization solutions for sectors such as government affairs, finance, healthcare, and industrial manufacturing. It promotes the deep integration of artificial intelligence (AI) and blockchain technologies, accelerating the standardization and large-scale application of AI in vertical industries.
As the lead organization for this standard, AsiaInfo Technologies has leveraged its self-developed "AISWare STC" and "AISWare AI² FL" to create an innovative privacy computing model that ensures "data availability without visibility and data retention within its domain".
AISWare STC boasts robust regulatory auditing capabilities, data collaboration and synergy, and data privacy protection. It supports the entire process of federated learning tasks with transparency, trustworthy traceability, and secure computation. AISWare AI² FL, through its universal middleware components, shields the differences in underlying blockchain technologies, and provides flexible and scalable on-chain federated learning services. This significantly lowers the barriers to entry for utilizing blockchain and federated learning technologies.
Additionally, AISWare STC and AISWare AI² FL offer comprehensive features such as data authorization management, on-chain recording of model training and inference processes, traceability of intermediate results, and task progress monitoring. This ensures controlled, traceable, and compliant circulation of data assets among relevant parties while fully protecting data privacy, thereby effectively enhancing the customers' long-term reliance on and sustained use of the platforms.
To date, AsiaInfo Technologies has integrated the federated learning and blockchain technologies in multiple commercial deployments, successfully applying them to business scenarios such as joint risk control, credit evaluation, and anti-fraud. By introducing these technologies, financial institutions have not only achieved the goal of "data retention within its domain, available of value release", but also gained efficient, personalized precision service capabilities, boosting user viscosity significantly.
Looking ahead, AsiaInfo Technologies will continue advancing cloud-based, modular, and scenario-based services of its platform capabilities, actively catering to a wider range of industries. It will help the clients across various sectors accelerate the discovery and release of the value of data elements, and construct a digital intelligence ecosystem that facilitates data circulation and privacy protection, driving the high-quality development of the digital economy.
