Vault Shield

Inspiration

Our inspiration came from witnessing how much unnecessary financial privacy people sacrifice every day. Whether applying for an apartment, a loan, or a visa, individuals are forced to hand over entire financial histories — exposing sensitive data far beyond what’s needed. We saw this as a broken model: invasive, risky, and outdated. The rise of zero-knowledge proofs and on-device AI presented a clear opportunity to redesign trust in digital finance—enabling verification without exposure, and giving users back control over their data.

What it does

VaultShield acts as a privacy firewall for financial identity. It offers two core modes of protection:

1. Proof without exposure – Using zero-knowledge proof circuits, users can generate verifiable “Financial Badges” (e.g., “Minimum Balance > $X”) that prove a fact without revealing underlying data like transaction history or exact balances.

2. Intelligent document protection – When documents must be shared, VaultShield uses on-device AI to perform context-aware redaction. It removes truly sensitive data (account numbers, CVV, spending patterns) while preserving necessary info, and can apply role-based viewing rules for different recipients.

All processing happens locally—your data never leaves your device unencrypted.

How we built it

We built VaultShield with a local-first, trustless architecture:

• Frontend: React Native + Expo for cross-platform mobile support, with a focus on intuitive UX for badge creation and document review.
• ZK Circuits: Custom-built using Circom and snarkjs for badge logic (balance, income, consistency proofs). Proofs are generated client-side.
• On-device AI: Integrated TensorFlow Lite and OCR engines (Tesseract) for local document analysis and smart redaction—no cloud processing.
• Privacy Orchestration Engine: Lightweight local ML model that learns user sharing patterns and flags anomalous data requests.
• Backend: Minimalist Node.js service acting as a relay for encrypted badge blobs and metadata—no access to user financial data.

Challenges we ran into

• Performance of on-device ZK proofs: Generating zero-knowledge proofs on mobile devices initially had high latency. We optimized circuits and used WebAssembly to bring generation time under 10 seconds.
• Accurate document understanding without cloud AI: Training a lightweight model to reliably distinguish “necessary” from “sensitive” data in diverse financial documents was difficult. We used a hybrid rules-based + ML approach to improve accuracy.
• UX for a new paradigm: Designing intuitive flows for concepts like “badges” and “privacy personas” required multiple iterations to make the technology accessible to non-technical users.

Accomplishments that we’re proud of

• Creating a fully functional prototype that generates and verifies ZK badges locally on a smartphone.
• Building an on-device redaction AI that works offline and removes sensitive data at the file-structure level, not just visually.
• Designing a system where users don’t have to trust us—their data never reaches our servers in plaintext.
• Winning early interest from proptech and fintech partners who see this as a solution to their own compliance and privacy challenges.

What we learned

• Zero-knowledge proofs are ready for practical privacy applications, but UX design is just as important as cryptographic correctness.
• On-device AI is powerful but requires careful model optimization to balance accuracy and performance.
• Privacy is a feature people value deeply once they understand the risk—the key is making it effortless.
• There’s significant appetite among institutions for privacy-preserving verification—if it can integrate smoothly into their workflows.

What’s next for VaultShield

• Expanded badge templates: Supporting credit score ranges, debt-to-income ratios, and custom logic for specific industries (immigration, auto leasing).
• Institution-side SDK: Allowing landlords, lenders, and employers to easily request and verify badges within their existing portals.
• Privacy network layer: Enabling users to earn reputation across applications without exposing identity, creating a portable, private financial profile.
• Beta launch: Pilot programs with property management companies and digital banks to refine real-world usability.

We believe VaultShield can redefine digital trust—turning financial verification from a privacy risk into a secure, user-controlled handshake.

Built With

• javascript
• next.js
• react
• snark
• tailwind
• tesseract
• typescript
• zk