Local Differential Privacy Explorer

Understanding Epsilon (ε) and the Privacy-Utility Tradeoff

Privacy Budget (Epsilon ε) 1.5
High Privacy
(More Noise)
High Accuracy
(Less Noise)

Current Probability:

82% chance to send True Data
18% chance to send Opposite Data (Noise)
Your Device (Local)

1. The Local Lottery

Before your data leaves your device, it plays a lottery based on the Epsilon setting above.

Medical Server (Cloud)

2. What They Receive

The server receives only the result of the lottery. It never sees your original choice or other local inputs.

Incoming Stream
👩‍💻
Analyst
Waiting for data...

Step 2: Collective Utility (1,000 Users)

Low ε protects privacy but creates highly inaccurate "NOISY TOTALS." "ESTIMATED TOTALS" uses the known lottery odds (ε) to reverse the noise.

TRUE TOTALS
80%
Never
20%
Dep
1. NOISY TOTALS
?
Never
?
Dep
2. ESTIMATED TOTALS
?
Never
?
Dep