臺灣博碩士論文加值系統

在大數據的時代，越來越多產業開始應用數據進行機器學習以提升企業效能或開發創新產品與服務。在進行機器學習的過程中，常面臨資料量不足、資料多元性不足的問題，使得訓練出來的模型結果不佳，但同業間基於競爭關係，不願意直接進行資料交換，部分產業則受個人資料隱私相關法規限制，無法進行客戶資料交換，使得企業難以發展更多元而有效的機器學習模型。
在過去幾年來，區塊鏈技術快速發展，其去中心化與不可篡改的特性，為企業帶來無限創新與想像力，目前已有許多企業廣泛接納區塊鏈技術，以區塊鏈技術帶來創新以及營運效能的提升，而區塊鏈衍生的「代幣經濟」更是將區塊鏈推向普及化的關鍵，給了人們未來邁向去中心化世界的想像，因此近年基於代幣經濟所衍生的區塊鏈應用蓬勃發展，也大幅提升了人們對於區塊鏈的接受度。因此，本研究提出了一個基於區塊鏈的數據分析模型共享協作平台，結合聯邦學習技術，讓企業組織間可不透過資料共享的形式進行模型建構的合作，並結合代幣經濟的概念，發展出一個新型態的企業協作模式。
聯邦學習 (Federated Learning) 是一種基於隱私保護的機器學習架構，可由各參與者以自己擁有的資料各自訓練模型後，整合成一個更強健的模型，即使資料不共享，也能進行模型開發的合作，提供各產業一個不做資料共享的情況下，進行機器學習共同合作的解決方案。而在原始的聯邦學習架構中，由於缺乏誘因，使得參與者參與意願較低落，因此本論文加入激勵機制的元素，計算各參與者在參與合作模型的貢獻程度，公平的分配獎勵，除了提升參與意願，也能夠更激勵參與者投入更多資源與心力以獲得更多獎勵，也能使模型能夠有更好的表現，獲得更大的效益。
激勵機制的建立與代幣經濟的結合，可為聯邦學習參與企業建立新的獲利模式，企業參與聯邦學習模型協作以獲取代幣作為獎勵，支付代幣來使用共享模型，而外部企業也可透過購買並支付代幣來獲取共享模型，而其支付金額依模型貢獻程度分配給參與者，讓參與者除了能夠獲取共享模型外也能夠額外獲取更多的報酬，期許能促進企業間更多的合作。
總而言之，本研究結合區塊鏈與聯邦學習技術，為企業間提供一個新型態的協作模式，並結合聯邦學習激勵機制與代幣經濟，為企業創造新的獲利模式，期許能活化企業間的交流，並建立起數據分析模型的供需生態圈，機器學習與數據科學的發展帶來更多的可能。

In the big data era, an increasing number of industries are beginning to apply machine learning using data to improve business efficiency or develop innovative products and services. However, during the machine learning process, there are often problems with insufficient data and a lack of data diversity, resulting in poor model results. Despite the competition, some industries are unwilling to exchange data directly, while others are limited by personal data privacy regulations and cannot exchange customer data. This makes it difficult for businesses to develop more diverse and effective machine-learning models.
In recent years, blockchain technology has experienced rapid development, and its decentralized and immutable characteristics have brought about infinite innovation and possibilities for enterprises. Numerous enterprises have already embraced blockchain technology to drive innovation and operational efficiency. The concept of a ”token economy” stemming from blockchain is crucial in popularizing the technology and inspiring people to envision a decentralized world. As a result, blockchain applications derived from the token economy have flourished in recent years, significantly boosting people’s acceptance of blockchain.
Therefore, this paper proposed a blockchain-based data analysis model sharing and collaboration platform, combined with federated learning technology, which allows organizations to collaborate on model construction without the need for data sharing. The concept of a token economy is also integrated to develop a new type of enterprise collaboration model.
Federated Learning has emerged as a privacy-preserving machine learning architecture that enables participants to train models using their data, culminating in the creation of a more robust model without the need for data sharing. This approach provides a compelling solution for diverse industries seeking to engage in collaborative machine-learning endeavors while upholding data privacy. However, the original Federated Learning architecture encounters a significant hurdle due to the lack of participant incentives, resulting in a diminished willingness to contribute. To address this issue, this thesis introduces an innovative incentive mechanism that quantifies each participant’s contribution to the collaborative model and facilitates the equitable distribution of rewards. By elevating willingness to participate and incentivizing participants to allocate additional resources and efforts in pursuit of greater rewards, this mechanism not only enhances overall participation but also augments model performance, leading to substantial benefits.
This paper explores the establishment of incentive mechanisms and their integration with the token economy, presenting novel profit models for enterprises engaging in Federated Learning. By participating in the collaborative model of Federated Learning, enterprises can earn tokens as rewards and utilize these tokens for accessing shared models. Furthermore, external enterprises can purchase and compensate tokens to acquire shared models, with the payment amount determined by the model’s contribution. The integration of incentive mechanisms and the token economy not only incentivizes enterprise participation but also fosters a dynamic ecosystem where value is exchanged through tokens, revolutionizing the profit landscape in the context of federated learning.

摘要 i
Abstract iii
Contents vi
List of Figures ix
List of Tables xi
Chapter 1 Introduction 1
1.1 Background............................... 1
1.2 Objective................................ 3
1.3 Paper Organization........................... 5
Chapter 2 Literature Review 6
2.1 DataPrivacy .............................. 6
2.1.1 Differential Privacy.......................... 7
2.1.2 Homomorphic Encryption ...................... 7
2.2 Distributed Machine Learning ..................... 8
2.2.1 Distributed Training.......................... 9
2.2.2 Federated Learning .......................... 9
2.3 Federated Learning Incentive Mechanism . . . . . . . . . . . . . . . 11
2.3.1 Data Quality.............................. 11
2.3.2 Data Quantity.............................. 12
2.4 Blockchain . . . . . . . . . . . . . . . 13
2.4.1 Smart Contract ............................ 13
2.4.2 Data and Model Sharing with Blockchain . . . . . . . . . . . . . . 14
2.4.3 Blockchain and Federated Learning ................. 15
2.4.4 Token Economy............................ 16
Chapter 3 Methodology 18
3.1 Design of Collaboration Platform Architecture . . . . . . . . . . . . 19
3.1.1 Design of Models Sharing Collaboration Mechanism . . . . . . . . 19
3.1.2 Design of Federated Learning Architecture. . . . . . . . . . . . . . 20
3.1.3 Design of Smart Contract....................... 21
3.1.4 Design of Incentive Mechanism ................... 23
3.2 Experiment............................... 25
3.2.1 Dataset................................. 26
3.2.2 Experiment Design .......................... 27
Chapter 4 Experiment Results 30
4.1 Experiment Setup............................ 30
4.2 Model Sharing Collaboration Platform. . . . . . . . . . . . . . . . . 31
4.2.1 FederatedLearningProcess...................... 31
4.2.2 Implementation of Collaboration Contract . . . . . . . . . . . . . . 31
4.2.3 ImplementationofFLTtokenContract . . . . . . . . . . . . . . . . 37
4.3 Effectiveness Experiment of the Platform . . . . . . . . . . . . . . . 37
4.3.1 Experiment of Federated Learning .................. 37
4.3.2 Experiment of Collaboration Mechanism. . . . . . . . . . . . . . . 41
Chapter 5 Conclusions 46
5.1 Conclusion ............................... 46
5.2 FutureWorks.............................. 48
References 49
Appendix A — Consortium Chain Transactions 53
A.1 Contract Deployment.......................... 53
A.2 Transactions .............................. 53
Appendix B — Public Chain Transactions 56
B.1 ContractDeployment.......................... 56
B.2 Transactions .............................. 57

[1] R. Michael and Lareina, “Mckinsey technology trends outlook 2022.” McKinsey & Company, 2022.
[2] H. Brendan McMahan, E. Moore, D. Ramage, S. Hampson, and B. Agüera y Ar- cas, “Communication-efficient learning of deep networks from decentralized data,” arXiv e-prints, pp. arXiv–1602, 2016.
[3] A. C. Yao, “Protocols for secure computations,” in 23rd annual symposium on foundations of computer science (sfcs 1982), pp. 160–164, IEEE, 1982.
[4] O. Goldreich, “Secure multi-party computation,” Manuscript. Preliminary version, vol. 78, no. 110, 1998.
[5] I.P.FellegiandA.B.Sunter,“Atheoryforrecordlinkage,”Journal of the American Statistical Association, vol. 64, no. 328, pp. 1183–1210, 1969.
[6] C. Dwork, “Differential privacy,” in Automata, Languages and Programming: 33rd International Colloquium, ICALP 2006, Venice, Italy, July 10-14, 2006, Proceedings, Part II 33, pp. 1–12, Springer, 2006.
[7] R. L. Rivest, L. Adleman, M. L. Dertouzos, et al., “On data banks and privacy homomorphisms,” Foundations of secure computation, vol. 4, no. 11, pp. 169–180, 1978.
[8] C. Gentry, “Fully homomorphic encryption using ideal lattices,” in Proceedings of the forty-first annual ACM symposium on Theory of computing, pp. 169–178, 2009.
[9] P. Paillier, “Public-key cryptosystems based on composite degree residuosity classes,” in Advances in Cryptology—EUROCRYPT'99: International Conference on the Theory and Application of Cryptographic Techniques Prague, Czech Republic, May 2–6, 1999 Proceedings 18, pp. 223–238, Springer, 1999.
[10] J. Verbraeken, M. Wolting, J. Katzy, J. Kloppenburg, T. Verbelen, and J. S. Rellermeyer, “A survey on distributed machine learning,” Acm computing surveys (csur), vol. 53, no. 2, pp. 1–33, 2020.
[11] Y. Zhan, J. Zhang, Z. Hong, L. Wu, P. Li, and S. Guo, “A survey of incentive mechanism design for federated learning,” IEEE Transactions on Emerging Topics in Computing, vol. 10, no. 2, pp. 1035–1044, 2021.
[12] R.Jia, D.Dao, B.Wang, F. A. Hubis, N. Hynes, N. M. Gürel, B. Li, C. Zhang, D. Song, and C. J. Spanos, “Towards efficient data valuation based on the shapley value,” in The 22nd International Conference on Artificial Intelligence and Statistics, pp. 1167–1176, PMLR, 2019.
[13] X.Yang, W. Tan, C.Peng, S.Xiang, and K.Niu, “Federated learning incentive mechanism design via enhanced shapley value method,” Wireless Communications and Mobile Computing, vol. 2022, 2022.
[14] Y.Zhan, P.Li, K.Wang, S.Guo, and Y.Xia, “Big data analytics by crowdlearning: Architecture and mechanism design,” IEEE Network, vol. 34, no. 3, pp. 143–147, 2020.
[15] Y.Zhan and J.Zhang, “Anincentivemechanismdesignforefficientedgelearningby deep reinforcement learning approach,” in IEEE INFOCOM 2020-IEEE Conference on Computer Communications, pp. 2489–2498, IEEE, 2020.
[16] S. Nakamoto, “Bitcoin: A peer-to-peer electronic cash system,” Decentralized business review, p. 21260, 2008.
[17] N. Szabo, “Smart contracts: building blocks for digital markets,” EXTROPY: The Journal of Transhumanist Thought,(16), vol. 18, no. 2, p. 28, 1996.
[18] V. Buterin et al., “A next-generation smart contract and decentralized application platform,” white paper, vol. 3, no. 37, pp. 2–1, 2014.
[19] K. R. Özyilmaz, M. Doğan, and A. Yurdakul, “Idmob: Iot data marketplace on blockchain,” in 2018 crypto valley conference on blockchain technology (CVCBT), pp. 11–19, IEEE, 2018.
[20] A. Marathe, K. Narayanan, A. Gupta, and P. Manoj, “Dinemmo: decentralized incentivization for enterprise marketplace models,” in 2018 IEEE 25th International Conference on High Performance Computing Workshops (HiPCW), pp. 95–100, IEEE, 2018.
[21] D. Justin and B. Harris, “Decentralized & collaborative ai on blockchain,” in Proceedings of the 2019 IEEE International Conference on Blockchain (Blockchain), Atlanta, GA, USA, pp. 14–17, 2019.
[22] H.Kim, J.Park, M.Bennis, and S.-L.Kim, “Blockchained on-device federated learning,” IEEE Communications Letters, vol. 24, no. 6, pp. 1279–1283, 2019.
[23] Y. Liu, Z. Ai, S. Sun, S. Zhang, Z. Liu, and H. Yu, “Fedcoin: A peer-to-peer payment system for federated learning,” in Federated Learning: Privacy and Incentive, pp. 125–138, Springer, 2020.
[24] X.FengandL.Chen, “Dataprivacyprotectionsharingstrategybasedonconsortium blockchain and federated learning,” in 2022 International Conference on Artificial Intelligence and Computer Information Technology (AICIT), pp. 1–4, IEEE, 2022.
[25] A. E. Kazdin and R. R. Bootzin, “The token economy: An evaluative review 1,” Journal of applied behavior analysis, vol. 5, no. 3, pp. 343–372, 1972.
[26] J. Benet and N. Greco, “Filecoin: A decentralized storage network,” Protoc. Labs, pp. 1–36, 2018.
[27] F. Vogelsteller and V. Buterin, “Erc-20: Token standard.” Ethereum Improvement Proposals, November 2015.
[28] S. Ellis, A. Juels, and S. Nazarov, “Chainlink: A decentralized oracle network,” Retrieved March, vol. 11, no. 2018, p. 1, 2017.
[29] Möbius, “Learning from imbalanced insurance data.”https://www.kaggle.com/ datasets/arashnic/imbalanced-data-practice, 2021.
[30] Hyperledger Foundation, “Hyperledger besu.” https://www.hyperledger.org/ use/besu, 2023.