Maven
A Decentralised AI-based KYC System
The problem Maven solves
Traditional Know Your Customer (KYC) processes are often redundant and inefficient. Users must repeatedly verify their identities across multiple platforms, even if a single organization controls multiple services offered by them. This is a pain point for both the organization as well as the user, as the user is stuck in time-consuming processes and suffers from redundant efforts, while the organization also suffers data privacy risks, becoming prime targets of cybersecurity attacks.
Our solution, which is a decentralized KYC system aided by blockchain, zero-knowledge proofs, and AI, solves these problems with the help of a one-time, reusable ID verification system:
- Single KYC Submission
- ZKPs ensure privacy
- Faster services, with no need to wait for manual KYC verification.
- Facial embeddings of the user, which are a good way to identify each user distinctly, get converted to hashes, further enhancing security.
Challenges we ran into
- MTCNN is a pre-trained model that occasionally provides poor cosine similarities between facial embeddings of the same person, thus a lot of experimentation had to be done with the pre-processing steps.
- Implementation of ZKP circuits on Circom required great effort since the technology was new to us.
- Verification of AADHAR was a crucial step. Still, it only has 2 ways of legal identification: firstly using the UIDAI API and secondly by analyzing the UIDAI signature in each AADHAR card. This is a challenge we are presently still trying to tackle.
Tracks Applied (1)
Ethereum Track
ETHIndia
Discussion
Builders also viewed
See more projects on Devfolio