🔒 Privacy-Preserving Cancer Detection with Fully Homomorphic Encryption (FHE)

This demo lets you upload a microscopic tissue image and get a “Cancerous” or “Not Cancerous” verdict—without ever revealing your raw data. Under the hood, we run the deep learning classifier from your image, encrypt it with fully homomorphic encryption (FHE), and run the classification entirely on ciphertext. That means all computation happens on encrypted values: the server never sees your clear-text image or features, yet still returns an accurate diagnosis in milliseconds.

Steps

  1. Generate your public/private keys
  2. Upload or select a reference cell image
  3. Generate & encrypt its embedding
  4. Set a decision threshold and click “Check”
  5. See the verdict (“Cancerous” vs. “Not Cancerous”)

Experience cutting-edge FHE in action: robust cancer screening with patient privacy built in.

Setup Phase: 🔐 Generate the FHE public and secret keys.

Choose a security level

Step 1: Upload or select a reference cell image

Step 2: Generate reference embedding.

Step 3: 🔒 Encrypt reference embedding using FHE.

Step 4: 🔒 Compute classification using the threshold under FHE.

Set the recognition threshold.

-2560 2560