📖 What is Microsoft Purview Data Classification?
Microsoft Purview Data Classification is the process of identifying and labeling data based on its sensitivity or type. It uses Sensitive Information Types (SITs) and trainable classifiers to automatically detect patterns, such as credit card numbers or medical records, across the M365 ecosystem.
"Understand the difference between a 'Sensitive Information Type' (pattern-based) and a 'Trainable Classifier' (machine learning-based) for the exam."
📚 Certification: Microsoft 365 Administrator (MS-102)
🔑 What are the Key Concepts of Microsoft Purview Data Classification?
- ▸ Sensitive Information Types (SITs) use regular expressions and checksums to identify structured data patterns like credit card numbers or Social Security numbers.
- ▸ Trainable Classifiers leverage machine learning to identify unstructured data based on context, such as legal contracts, resumes, or harassment-related content.
- ▸ Auto-labeling policies allow administrators to automatically apply sensitivity labels to content that matches specific classification criteria across M365 services.
- ▸ Confidence levels determine the certainty of a match, allowing admins to reduce false positives by requiring higher thresholds for automatic classification.
- ▸ Data classification serves as the foundation for Data Loss Prevention (DLP) policies, enabling the system to block or encrypt sensitive information.
🎯 How does Microsoft Purview Data Classification appear on the MS-102 Exam?
You may be asked to determine whether a Sensitive Information Type or a Trainable Classifier is appropriate for detecting unstructured documents, such as internal project proposals.
A scenario might describe a requirement to automatically protect files containing PII across SharePoint and OneDrive; you must identify the correct auto-labeling configuration.
Expect questions about adjusting confidence levels to resolve a situation where too many non-sensitive documents are being incorrectly flagged as sensitive.
❓ Frequently Asked Questions
When should I use a Trainable Classifier instead of a Sensitive Information Type?
Use SITs for structured data with predictable patterns, such as government IDs. Use Trainable Classifiers for unstructured data where the meaning depends on the overall context of the document.
How does the confidence level impact the classification process?
Confidence levels define how strongly a pattern must match to be flagged. Higher levels reduce false positives but may miss some valid sensitive data if the pattern is imperfect.
Does classification automatically encrypt the data?
No, classification identifies the data. Encryption occurs only if the classification triggers a Sensitivity Label that has encryption settings configured within its specific policy.