EpsiLint: A Linter for Catching Common Mistakes in Differential Privacy Code
NIST 800-226 · W3C DP · van der Veen · DiffMu — YAML-driven policy engine (static offline build)
Python source code
Deployment context
Dataset size (n)
Trust model
Central (trusted curator)
Local (on-device noise)
Distributed / shuffle
Not specified
Data sensitivity
High (medical, financial, PII)
Medium (behavioral, usage)
Low (aggregate, non-PII)
Unit of privacy
User-level
Event-level
Not specified
Training mode
Not DP-SGD
DP-SGD
Federated DP
Run privacy analysis
Clear
Export rules (YAML)
Export rules (JSON)
Import custom rules
EpsiLint v3.0 — Policy engine driven by NIST SP 800-226, W3C DP Guidance, van der Veen et al. (2018), and DiffMu type system (static offline build)