Chang, E.Y.E.Y.ChangSHIH-WEI LIAOLiu, C.-T.C.-T.LiuLin, W.-C.W.-C.LinLiao, P.-W.P.-W.LiaoFu, W.-K.W.-K.FuMei, C.-H.C.-H.Mei2020-05-042020-05-042019https://scholars.lib.ntu.edu.tw/handle/123456789/489697https://www.scopus.com/inward/record.uri?eid=2-s2.0-85062187518&doi=10.1109%2fAIVR.2018.00037&partnerID=40&md5=f0dac7b3355e36d97a0347dfd4b3cf3bThis paper presents requirements to DeepLinQ and its architecture. DeepLinQ proposes a multi-layer blockchain architecture to improve flexibility, accountability, and scalability through on-demand queries, proxy appointment, subgroup signatures, granular access control, and smart contracts in order to support privacy-preserving distributed data sharing. In this data-driven AI era where big data is the prerequisite for training an effective deep learning model, DeepLinQ provides a trusted infrastructure to enable training data collection in a privacy-preserved way. This paper uses healthcare data sharing as an application example to illustrate key properties and design of DeepLinQ. © 2018 IEEE.Blockchain; Consensus protocol; Ledger system; Privacy preserving; Smart contract[SDGs]SDG9Access control; Artificial intelligence; Blockchain; Deep learning; Virtual reality; Application examples; Consensus protocols; Distributed data; ITS architecture; Learning models; Ledger system; Privacy preserving; Training data; Data privacyconference paper10.1109/AIVR.2018.000372-s2.0-85062187518