Data privacy is a primary concern of the Matrix AI Network.
Data is fundamental to the AI industry, particularly as it pertains to AI model training. However, as data transparency becomes an increasingly important topic in the big data era, the issue of how to avoid data leakage comes into focus. This is an issue that should be of particular concern to the blockchain industry. This article introduces how the Matrix AI Network protects data privacy by underpinning its Matrix 2.0 platform with distributed storage solutions during data upload, as well as by using federal learning, homomorphic encryption and secure multi-party computation during data computation.
As data sharing and collaboration come to prevalence in today’s big data era, traditional blockchain solutions designed to openly store data cannot meet the growing needs of data protection. As a result, the blockchain industry will focus on data protection while supporting various forms of data attribution to allow a distinction between the right to access, use and own data.
The Matrix AI Network is building a blockchain-based AI economy where recording data directly to an encrypted distributed ledger ensures exclusive attribution of data ownership rights and prevents data from being furtively reproduced. Such a distributed storage system can ensure data privacy, while bridging data islands to create fertile new opportunities for big data to evolve; including collaboration and profit splits on data generation, modeling, applications.
To efficiently protect data privacy, Matrix AI Network has introduced the following three major technologies:
Federated learning is a distributed training method for machine learning (ML). Each device only processes a portion of the ML training task, and the training results are integrated later. With this technology, no individual device can have access to complete data so that data privacy can be properly protected.
Homomorphic encryption is a form of encryption that allows computation on ciphertexts, generating an encrypted result which, when decrypted, matches the result of the operations as if they had been performed on the plaintext. With this technology, users asking for computing power will no longer open plaintext to cloud server, which can efficiently avoid data leakage.
Secure multi-party computation (SMC）
SMC is a subfield of cryptography with the goal of creating methods for parties to jointly compute a function over their inputs while keeping those inputs private and preserving independence and computational accuracy. Such a framework has been widely used in fields involved with sensitive information such as customer behavior, identification, credit inquiry and so on. On the Matrix 2.0 AI-powered blockchain platform, SMC ensures data scientists jointly process separately-stored data and except computational results, inputs cannot open to all nodes.
As high-quality data can better train AI models so as to efficiently advance the development of AI technology, the Matrix AI Network endeavors to protect data privacy.