Differential Privacy Epsilon Budget Calculator
Track a user-supplied Differential Privacy epsilon budget using basic sequential composition. Enter prior spend, a per-release epsilon and planned release count to estimate the remaining budget; this tool does not select a privacy threshold or privacy mechanism.
Result
Epsilon remaining after planned releases
Projected total epsilon spent
Projected budget used
Additional equal-size releases after plan
How to use this tool
- Enter the total epsilon budget for the dataset or privacy accounting scope.
- Enter epsilon already spent, the epsilon charged to each planned release and the number of releases.
- Review projected spend and remaining epsilon. A negative remaining value means the planned releases exceed the entered budget under basic composition.
Assumptions
- Uses basic additive sequential composition for pure Differential Privacy. It does not apply advanced composition, parallel composition, subsampling amplification or an RDP accountant.
- Delta for approximate DP is not modeled. If the mechanisms use delta or mixed accounting methods, this arithmetic is not a complete privacy accountant.
- The tool only tracks values you enter; it does not determine whether an epsilon, delta, mechanism or release schedule is appropriate for a dataset or policy.
Sources & reference data
Technical details: what the fields mean
- Total epsilon budget
epsilonBudget - Maximum epsilon budget for the accounting scope, as determined by your privacy policy or analysis.
- Epsilon already spent
epsilonSpent - Previously composed epsilon spend for the same scope and accounting method.
- Epsilon per planned release
epsilonPerRelease - Epsilon charged to each upcoming release; this value is user-supplied and is not derived from a mechanism.
- Number of planned releases
plannedReleaseCount - Number of equal-epsilon releases to include in the projection.
Methodology
Under basic sequential composition for pure DP, projected epsilon spent = prior epsilon + epsilon per release × planned release count. Remaining epsilon = total budget − projected spend. Additional equal-size releases after the plan = floor(max(0, remaining epsilon) ÷ epsilon per release).
Worked example
Four planned releases at epsilon 0.35 each
Inputs: Total epsilon budget: 5 epsilon, Epsilon already spent: 2.4 epsilon, Epsilon per planned release: 0.35 epsilon/release, Number of planned releases: 4 releases
Result: Projected spend is 3.8 epsilon; 1.2 epsilon remains. At 0.35 epsilon per release, three more equal-size releases would fit under basic composition.
Practical guidance
- Keep one consistent accounting scope and record the epsilon charged by each release before composing totals.
- Review whether your mechanisms use pure DP or approximate DP (epsilon, delta) and whether composition or amplification assumptions apply before relying on a simple sum.
Common mistakes
- Treating epsilon as a universal privacy score with a recommended threshold; acceptable budgets depend on the mechanism, data, release schedule and policy.
- Adding epsilon values from different datasets or accounting scopes as if they belonged to one budget.
FAQs
Does this calculate delta or advanced composition?
No. It uses basic additive composition for epsilon only. It does not calculate delta, advanced composition, subsampling amplification or RDP.
Does a positive remaining epsilon mean the release is private enough?
No. It only means the entered arithmetic is within the entered budget. Selecting mechanisms and thresholds requires a complete privacy analysis and policy.
Last reviewed: 2026-10-18
A simple epsilon ledger, not a Differential Privacy accountant. It does not inspect mechanisms, compute epsilon from data/noise, model delta, or provide security, legal or privacy-policy advice.
This tool provides an indicative estimate, based on the assumptions stated above.
Theme and tool preferences, along with saved and recently used lists, are stored only in this browser.