A client-side LLM agent automating the form submission journey while providing proof of data security and form submission via Mina protocol
This project streamlines form-filling by using AI to map responses and autofill data. It prioritizes privacy with zero-knowledge proofs, ensuring data isn't leaked during the process. Users have the option to reveal certain information for future convenience. The system is designed to be tamper-proof and easy to verify.
In our project, we harnessed the combined power of Worldcoin, Gaia node, and the Mina protocol to create a secure and efficient form-filling solution. Worldcoin's SDK enabled us to verify human requests, while the Gaia node's local LLM model intelligently mapped user responses to form data. The Mina protocol ensured data privacy through zero-knowledge proofs, preventing leaks during form submission.
We integrated our LLM with a Selenium backend, allowing it to directly populate forms on government websites. After the bot fills all values, users can choose to publicly reveal certain details for future auto-filling.
Once the form is filled and the user's choice is made, we send the input to our Mina contract for on-chain verification. Subsequently, we update the Merkle Mapping in our system to keep the FormVerifier contract's state current. If users opt to reveal data, we send it along with the commitment to avail, making it available for future form-filling.
This approach ensures our system's tamper-proof nature and simplifies LLM verification by focusing on the final hashed values during submission.