Differentially Private Tetris

Differential Privacy (DP) is a mathematical framework that guarantees privacy by adding noise to data. Think of it like wearing a mask - the more noise you add, the harder it is to identify individual features, but you can still understand what's going on!

The privacy parameter epsilon controls how much noise we add:

• Lower epsilon = More Privacy = More Noise
• Higher epsilon = Less Privacy = Less Noise

* Also we'll be maintaining a leaderboard for this based on screenshots!

Randomized Response is a simple yet powerful DP technique. Imagine someone asking you a personal yes/no question:

• Step 1: Flip a coin
• Step 2: If HEADS (p%), Tell the truth
• Step 3: If TAILS (1-p%), Flip again and:
• Second HEADS: Say "Yes"
• Second TAILS: Say "No"

This gives you "plausible deniability" - nobody knows if your answer came from the coin or the truth!

The probability p is related to epsilon:

• epsilon = 0: p = 0% (complete privacy)
• epsilon = log(3): p = 50% (some privacy)

We apply Randomized Response to each square of the Tetris block:

• For each cell position (Using RR):
• Show true block?
• If showing but not true, what color?


• Privacy Budget:
• Each peek costs epsilon
• Higher epsilon = Clearer view
• Lower epsilon = More uncertainty

Try moving the epsilon slider below to see how privacy affects block visibility. Watch how individual squares might change or disappear, making it harder to determine the true block shape!

In this game, we protect the privacy of each falling Tetris block using Differential Privacy (DP). The more privacy you want, the harder it gets to see the blocks clearly!

Higher Epsilon = Less Privacy, Clearer Blocks
Lower Epsilon = More Privacy, Harder to See