Privacy Impact Assessments and Tools
Privacy Impact Assessment (PIA / DPIA)
A Data Protection Impact Assessment (DPIA) is a process designed to help systematically analyze, identify, and minimize the data protection risks of a project.
When is it required?
- Large-scale processing of highly sensitive data.
- Systematic monitoring of publicly accessible areas.
- Processing using new technologies (e.g., AI/ML).
LINDDUN Threat Modeling
LINDDUN is a privacy threat modeling methodology (similar to STRIDE for security). It stands for:
- Linkability
- Identifiability
- Non-repudiation
- Detectability
- Disclosure of information
- Unawareness
- Non-compliance
Applying LINDDUN
- Model the System: Create Data Flow Diagrams (DFDs).
- Map Threats: Apply the LINDDUN categories to the DFD elements (Entities, Data Flows, Data Stores, Processes).
- Analyze: Assess the risk and impact of each threat.
- Mitigate: Apply privacy-enhancing technologies and controls.
Privacy Tooling Ecosystem
- Google Differential Privacy Library: C++, Java, and Go libraries for generating differentially private statistics.
- Diffprivlib: IBM's Differential Privacy Library for Python, extending scikit-learn.
- OpenMined / PySyft: A Python library for secure, private Deep Learning (Federated Learning, SMPC, DP).
- ARX Data Anonymization Tool: An open-source tool for anonymizing sensitive personal data using models like k-anonymity.