Privacy Budget
A quantitative measure (epsilon, ε) of the total privacy loss accumulated through repeated queries on privacy-protected data.
The Privacy Budget (epsilon) limits how many queries on protected data are possible before privacy erodes – a finite allowance for each data source.
Explanation
Each DP query consumes part of the budget. Once the budget is exhausted, no further queries are possible without additional privacy loss. Composition theorems describe how budget adds across queries.
Marketing Relevance
Analytics teams must manage their privacy budget: Too many reports consume it. Strategic prioritization of queries becomes necessary.
Example
A company sets ε=1.0 as annual privacy budget. After 100 marketing queries at ε=0.01 each, the budget is exhausted – further queries require new data release.
Common Pitfalls
Budget management is operationally complex. Too strict a budget blocks analytics. Composition isn't always tight – can be conservative.
Origin & History
Dwork et al. formalized the privacy budget concept in the Differential Privacy framework in 2006. Rényi DP and Gaussian DP (2017+) improved composition bounds for tighter budgets.
Comparisons & Differences
Privacy Budget vs. Differential Privacy
Differential Privacy is the framework; Privacy Budget is the quantitative measure within it that controls consumption.
Privacy Budget vs. Rate Limiting
Rate Limiting limits requests per time; Privacy Budget limits cumulative privacy loss across all queries.