IRIS 1

From Damage Control to Damage Prevention

What it does

Iris transforms the way people and businesses censor sensitive data in mass.

Millions of users share videos on social media everyday. However, the current system struggles to detect and remove sensitive information in real-time. Videos containing passwords and personally identifiable information (PII) like addresses, credit card details and IDs are frequently taken down only after user reports, which can be too late since the information has already been leaked.

This is where Iris steps in. By analyzing visual data for any sensitive data leaks before the video is published, we can solve the problem at its source.

Here's how it works: Clients upload their videos through an API request. These videos are encrypted in Firebase, processed on GCP, and analyzed for sensitive info. Our internal database of sensitive info is collected only on the consent of our users, who receive a decrease in their subscription fee. This data will then be deleted after a 30 day retention period. A report is generated, allowing clients to customize blur settings. Finally, IRIS returns a censored video with requested PII blurred.

We use Google Gemini 1.5 Pro to analyze any potential information leak. Additionally, we use RAG to retrieve any matches to our internal sensitive ino database. Google Gemini allows us to detect sensitive data from a semantic perspective. So more complex cases of PII and password leakage like mirrored or occluded texts can also be detected.

Built with

  • Web/Chrome
  • Firebase

Team

By

Iris

From

United States