SBTLF：一種使用基於區塊鍊和基於令牌的激勵化保護協作機器學習的新方法

人工智能（AI）繁榮的主要問題之一是用於培訓AI的數據的隱私。隨著國家現在熱衷於構建更多量身定制的AI和大型語言模型（LLM），我們如何使用和保護數據變得至關重要。

As nations strive to develop more sophisticated Artificial Intelligence (AI) and large language models (LLM), the privacy of the data used to train these models becomes a primary concern. To address this issue, researchers at the Indian Institute of Technology (IIT) Kharagpur have proposed a novel approach that leverages blockchain technology to facilitate secure data sharing for AI training.

隨著各國努力開發更複雜的人工智能（AI）和大型語言模型（LLM），用於訓練這些模型的數據的隱私成為主要問題。為了解決這個問題，印度技術學院（IIT）Kharagpur的研究人員提出了一種新穎的方法，該方法利用區塊鏈技術來促進安全數據共享進行AI培訓。

Their research focuses on enhancing Federated Learning (FL), a type of machine-learning model, by integrating it with blockchain technology and other privacy measures. This combined approach aims to make machine learning models more secure and equitable for all participants.

他們的研究重點是增強聯合學習（FL），即一種機器學習模型，通過將其與區塊鏈技術和其他隱私措施集成在一起。這種合併的方法旨在使機器學習模型對所有參與者更加安全和公平。

At its core, the Federated Learning model is designed to be trained across multiple devices without requiring the transfer of raw data. Similar to how a large project can be divided into smaller tasks to be worked on by individual personal computers, FL enables individual devices (such as smartphones and PCs) to contribute to the training of an AI model without having to share the private data they hold.

從本質上講，聯合學習模型旨在在不需要原始數據傳輸的情況下對多個設備進行培訓。類似於如何將大型項目分為較小的任務，可以由個人計算機來處理，FL使單個設備（例如智能手機和PC）可以為培訓AI模型做出貢獻，而無需共享他們持有的私人數據。

In FL, each individual device or ‘node’ trains a standard model on its local data and shares only the results—not the actual data—with a central server. To ensure privacy, the data is typically encoded using Local Differential Privacy (LDP). This well-established privacy model adds a layer of 'noise' to the data before it is shared, preserving the privacy of individual information.

在FL中，每個單獨的設備或“節點”都在其本地數據上訓練標準模型，並且僅與中央服務器共享結果（而不是實際數據）。為了確保隱私，通常使用局部微分隱私（LDP）對數據進行編碼。這個良好的隱私模型在共享數據之前為數據添加了一層“噪聲”，並保留了個人信息的隱私。

In the present study, the researchers introduce a new approach called SBTLF (Secure Blockchain-Based Tokenized LDP Federated Learning). They realized that by combining blockchain technology with federated learning and adding a system of tokens and Local Differential Privacy, they could train machine learning models more securely and efficiently.

在本研究中，研究人員介紹了一種稱為SBTLF的新方法（基於安全區塊鏈的LDP聯合學習）。他們意識到，通過將區塊鏈技術與聯合學習結合併添加一個令牌和當地差異隱私系統，他們可以更安全，有效地訓練機器學習模型。

Blockchain technology, which is better known for its association with cryptocurrencies such as Bitcoin, enables users to share data stored as tokens in a ledger in a decentralized and secure manner. For this purpose, the researchers employed HyperLedger Fabric, a specialized blockchain framework designed to securely manage data and transactions among different parties in a federated learning environment. It serves as a distributed ledger that records all the local nodes' updates and helps ensure the transparency of the entire process. This adds a layer of security by preventing any single point of failure, which could occur if only one system were to control the entire process.

區塊鏈技術以與比特幣這樣的加密貨幣關聯而聞名，它使用戶能夠以分散且安全的方式共享以代幣為代幣的數據。為此，研究人員採用了Hyperledger Fabric，這是一種專門的區塊鏈框架，旨在安全地管理聯合學習環境中不同各方之間的數據和交易。它充當分佈式分類帳，記錄所有本地節點的更新，並有助於確保整個過程的透明度。這通過防止任何單點故障來增加安全性，如果只有一個系統控制整個過程，則可能發生。

Furthermore, the researchers propose the use of a token-based incentive system to enable fair data sharing and utilization. One major challenge encountered with the FL model is ensuring that participants are willing to share their data and that they follow the protocols. In some systems, participants could send heavily obscured, and therefore less helpful, data and still access the benefits of the learning model as others who contributed more meaningful learning experiences.

此外，研究人員建議使用基於令牌的激勵系統來實現公平的數據共享和利用。 FL模型遇到的一個主要挑戰是確保參與者願意共享他們的數據並遵循協議。在某些系統中，參與者可能會發送大量遮蓋的數據，因此數據的數據較小，並且仍然可以訪問學習模型的好處，因為其他人貢獻了更有意義的學習經驗。

The SBTLF method tackles this problem by introducing token-based incentives that are linked to the privacy parameter. Participants earn tokens by sharing less obscured (less noise-added) data. Clearer (less obscured) data earns more tokens. These tokens can then be used to access global model updates and to reward participants who contribute more valuable data. Additionally, by encrypting the model updates and using blockchain for security, the risks associated with security attacks and a single point of failure are drastically reduced.

SBTLF方法通過引入與隱私參數相關的基於令牌的激勵措施來解決此問題。參與者通過共享較不遮蓋（噪音較少）的數據來賺取令牌。更清晰（不太模糊）的數據可獲得更多的令牌。然後，這些令牌可用於訪問全球模型更新，並獎勵貢獻更有價值數據的參與者。此外，通過加密模型更新並使用區塊鏈進行安全性，與安全攻擊和單個故障相關的風險大大減少了。

While this research opens up many exciting possibilities for secure collaborative machine learning, setting up a blockchain-based federated learning system can be technically complex and may require substantial computational resources. Moreover, introducing token-based incentives, which essentially involves a form of trading, may also require policy-level interventions.

儘管這項研究為安全的協作機器學習打開了許多令人興奮的可能性，但建立基於區塊鏈的聯合學習系統在技術上可能很複雜，並且可能需要大量的計算資源。此外，引入基於令牌的激勵措施本質上涉及一種交易形式，也可能需要政策級別的干預措施。

Nevertheless, the research provides a promising approach to making federated learning more secure and efficient. It introduces privacy tools and a clever reward system to ensure fair participation. By exploring ways to ease implementation and enhance resistance to different types of attacks, the research can ultimately pave the way for federated learning models to learn from decentralized data in a secure and effective manner. As the field continues to grow, integrating blockchain and privacy measures could set high standards for global data security and collaborative learning.

然而，這項研究提供了一種有希望的方法，使聯盟學習更加安全有效。它引入了隱私工具和巧妙的獎勵系統，以確保公平參與。通過探索緩解實施並增強對不同類型攻擊的抵抗力的方法，該研究最終可以為聯合學習模型以安全有效的方式從分散數據中學習。隨著領域的不斷增長，整合區塊鍊和隱私措施可能會為全球數據安全和協作學習設定高標準。

This research news was partly generated using artificial intelligence and edited by an editor at Research Matters

該研究新聞部分使用人工智能生成，並由研究事項編輯編輯