Restricted Verifier: Using ML/AI to Limit Data Disclosure

In the context of decentralized identities and self-sovereign identities, concern over data disclosure rises up. Legal regulations and policies are implemented to limit data misuse. However, users who are information holders do not have access to verify the data request from the service provider. To address this challenge, this work propose to leverage machine learning and artificial intelligence technology to help information holders decide if service providers only request strictly necessary data. Specifically, this work develops a restricted verifier prototype with two key components. One is on-chain smart contract to facilitate communication between entities. Another one is o↵-chain checker which consists of a machine learning model to determine the necessity of a data request. The system’s e↵ectiveness and efficiency are evaluated by testing both unsupervised and supervised learning models on synthetically generated datasets. In capability and performance tests, the Random Forest algorithm demonstrates exceptional performance in detecting excessive data requests. Furthermore, the restricted verifier prototype introduced minimal latency, indicating its application in the real world.