Skip to content
← Back to projects

PrivacyLens (privacyaudit)

An open-source 5-point AI privacy auditing framework and automated governance gate for Azure ML and Azure OpenAI deployments.

  • Python
  • Azure ML
  • Azure OpenAI
  • Scikit-Learn
  • PyPI
  • Cybersecurity
View liveView code

Overview

PrivacyLens (privacyaudit) is an enterprise-grade AI privacy evaluation framework designed to detect data memorization, PII leakage, and inference vulnerabilities in machine learning models and fine-tuned LLM endpoints.

Key Features

  • Membership Inference Attacks (MIA): Shadow model evaluation to measure whether specific training records can be reconstructed.
  • PII Leakage Scanner: Automated regex and entropy analysis for SSNs, emails, credit cards, and API secrets.
  • Azure ML Governance Gate: Drop-in AzureMLAuditStep for Azure Machine Learning pipelines that automatically halts non-compliant model registrations.
  • Azure OpenAI Fine-Tuning Audits: Pre-production prompt injection and prefix probing suites for fine-tuned LLM deployments.