DataTerminology.comNEW
An attribute containing personal, confidential, or protected information requiring special handling.
DESCRIPTION
Sensitive attributes are data fields that can reveal personal identity, protected characteristics, or confidential information. In privacy and fairness contexts, they include special categories under regulations like GDPR and protected characteristics in anti-discrimination laws. Proper handling involves consent, minimization, access control, and, when appropriate, anonymization or pseudonymization. In responsible AI, sensitive features inform fairness assessments without being used for discriminatory decisions.
ALTERNATIVE DEFINITIONS
GDPR-Info
Processing of personal data revealing racial or ethnic origin, political opinions, religious or philosophical beliefs, or trade union membership, and the processing of genetic data, biometric data for the purpose of uniquely identifying a natural person, data concerning health or data concerning a natural person’s sex life or sexual orientation shall be prohibited.
Defines special categories of personal data requiring heightened protection.
Fairlearn
Sensitive features are the features used to define the groups whose fairness we are interested in, such as sex or race.
Frames sensitive attributes for fairness assessment groups.
ISO 27701
ISO/IEC 27701 extends ISO/IEC 27001 and ISO/IEC 27002 to establish requirements for a privacy information management system, including controls for processing personal data.
Highlights governance controls for sensitive personal data processing.
RELATED TERMS
Protected Attribute Special Category Data PII PHI Fairness Bias Mitigation Anonymization Data Masking Attribute-Based Access Control Consent Management
COMPETING TERMS
Non-sensitive Attribute
KNOWLEDGE AREAS
Data Privacy & Compliance Data Security AI Governance Data Management
TARGET AUDIENCE
Data Privacy Officer Data Governance Manager Chief Information Security Officer Data Steward AI Engineer Data Scientist
EXTERNAL SOURCES
- Wikipedia - Personal data#Special category data
- European Union - Official GDPR regulation text
- NIST - Privacy Framework for data governance
- ICO - Guidance on UK GDPR special categories
- Google Cloud - Privacy and data protection overview
TOP QUESTIONS
What makes an attribute sensitive in data governance?
Sensitivity arises from potential harm or discrimination if exposed, legal obligations, and confidentiality requirements. Examples include health, sexual orientation, and biometric identifiers.
Category: Requirements
How should Sensitive Attributes be handled in machine learning?
Use them to assess fairness and bias but avoid discriminatory usage. Apply access controls, anonymization, and governance approvals.
Category: BestPractice
What are common mistakes with Sensitive Attributes?
Storing them unnecessarily, inadequate access controls, and using them as predictive features without governance can breach laws and ethics.
Category: Mistakes
What is the risk of mishandling Sensitive Attributes?
Legal penalties, reputational damage, and harm to individuals can result. Regulatory non-compliance may trigger audits, fines, or injunctions.
Category: Risk
How do GDPR special categories relate to Sensitive Attributes?
Special categories are legally recognized sensitive data requiring stricter controls and lawful bases, including explicit consent or specific exemptions.
Category: Opinion
How to minimize Sensitive Attributes in data pipelines?
Apply data minimization, pseudonymization, and masking. Use privacy impact assessments and retention policies.
Category: Management
Can Sensitive Attributes be used for auditing fairness?
Yes, with appropriate governance and consent, sensitive attributes define groups to measure disparate impact.
Category: Application
Where to start learning compliance for Sensitive Attributes?
Review GDPR, NIST Privacy Framework, and regulator guidance like ICO, then establish internal privacy policies.
Category: Requirements
What is a competing term for Sensitive Attribute?
Non-sensitive attribute refers to data not requiring heightened protection, but context may still impose controls.
Category: Alternatives