|
Work item:
|
Y.FML-AISF
|
|
Subject/title:
|
Federated machine learning-based artificial intelligence software framework with blockchain-driven edge data sharing
|
|
Status:
|
Under study
|
|
Approval process:
|
AAP
|
|
Type of work item:
|
Recommendation
|
|
Version:
|
New
|
|
Equivalent number:
|
-
|
|
Timing:
|
2027-Q4 (Medium priority)
|
|
Liaison:
|
ITU-T SG11, SG13, SG17, SG21
|
|
Supporting members:
|
ETRI, China Unicom, China Information Communication Technologies Group(CICT), Daejeon University (Korea), Algeria
|
|
Summary:
|
In order to enable distributed training and deployment of AI models while preserving privacy, enhancing trust, and promoting interoperability across diverse edge devices, this Recommendation employs federated machine learning, remaining raw data on local devices, reducing privacy leakage and network overhead, whereas uses blockchain technologies providing immutable records, smart contract–based governance, and secure edge-to-edge data sharing.
In traditional FML-based services, FML coordinators and participants exchange information directly, as specified in Recommendation ITU-T F.748.35. However, traditional FML-based services face several challenges in distributed or decentralized environments due to the lack of trust among multiple participants, regardless of whether such environments are considered trusted or untrusted. In such cases, distributed ledger technologies (DLT) provide inherent advantages for data management and sharing, including peer-to-peer communication, decentralization, immutability, openness, crowd consensus, and support for smart contract automation. The convergence of FML and DLT can therefore provide significant benefits for FML-based services. To address these challenges, it is important that FML supports distributed model training and inference across heterogeneous edge environments within the artificial intelligence software framework (AISF).
|
|
Comment:
|
-
|
|
Reference(s):
|
|
|
Historic references:
|
|
Contact(s):
|
|
| ITU-T A.5 justification(s): |
|
|
|
|
First registration in the WP:
2025-10-09 16:40:13
|
|
Last update:
2025-10-13 11:42:38
|
|
