Module 13 · Project 7 · Due the date on Canvas · Ethics, Privacy & Dashboard Detective · adaptive competency DV21
Small-cell & differencing-attack simulator
Mission progress0%
Begin by evaluating base-rate fairness and Simpson's paradox.
Module 13 · Step 3 of 9 · Guided10–12 min
Why this lab Block small-cell and differencing attacks before publishing a dashboard.
Granular filters can expose individual identities when cells drop below \(n < 5\) or when subtraction queries isolate a single record. Simulate queries, detect leaks, and apply cell suppression and differential noise.
- Removing names is not enough. In a cell of three students, anyone who knows who those students are can read their outcome off a 67% tile; they can be re-identified without a single name on screen.
- Protect at the display layer. Suppress small cells where results are published, not in the data: never delete records or subgroups from the analysis dataset, because removing people destroys evidence and biases every summary computed from what is left. Keep the data, suppress the display, and document the threshold.
Privacy simulator loading…
Differential privacy: the noise-for-protection trade
Suppression hides small cells; differential privacy protects even the published ones. Trade accuracy for protection with the privacy parameter \(\varepsilon\) and see why an attacker who can repeat queries still loses — as long as the query budget holds.
The differential-privacy lab needs JavaScript.