CIpher: Conceal Identity

CIpher - a Chrome extension that utilizes on-device AI to automatically detect and blur sensitive information in real-time, keeping your screen private during calls, recordings, and everyday browsing.

Comment

Inspiration

As remote collaboration increases, so do the risks of unintentional data exposure during screen sharing, live demos, and everyday browsing.
Cipher was created to explore the privacy-preserving applications of on-device AI, demonstrating how local language models, such as Gemini Nano, can securely enhance personal data protection without transferring user content to external servers.

What it does

Cipher is a Chrome extension that:

  • Leverages Gemini Nano for PII detection
  • Detects sensitive information such as emails, passwords, contact data, SSNs, credit card numbers, and API keys
  • Automatically blurs this information in real time on any webpage
  • Performs 100% on-device inference. No data leaves the browser
  • Displays a dynamic privacy score and contextual risk insights

How we built it

We developed ScreenGuard as a Chrome extension using Manifest V3 and the Chrome Prompt API (Gemini Nano) for on-device AI processing. The system continuously scans visible text nodes in the DOM, identifies PII, and overlays blur masks precisely over detected spans in real time.

Challenges we ran into

Area Challenge
AI Output Reliability Ensuring the model returned strict JSON with exact text indices
Precision Masking Translating model offsets into DOM ranges across dynamic page layouts
Scrolling & Layout Shifts Maintaining mask alignment during scroll/resize events
Model Initialization Handling on-device model load latency & user-activation requirements
Performance Balancing continuous scanning with browser performance constraints

Accomplishments that we're proud of

  • Built AI protection that runs locally, so your data never leaves your device.
  • Created privacy and safety scores that are simple to understand.
  • Designed an interface that feels safe, not intrusive, even when browsing anywhere.

What we learned

Through building this project, we gained hands-on experience with:

  • Chrome’s new on-device AI capabilities (Prompt API / Gemini Nano)
  • Enforcing structured model outputs for deterministic UI logic
  • Advanced DOM manipulation using text ranges and bounding rectangles
  • Designing privacy solutions that avoid cloud dependencies entirely
  • Performance strategies for continuous real-time browser extensions

What's next for CIpher

  • ML-guided false-positive reduction
  • Expand to images.
  • Make scanning even faster and more lightweight in the browser.
  • Exportable privacy reports.

Built With

Share this project:

Updates