Generalization and Suppression
What you'll learn: You'll discover two fundamental data transformation techniques that reduce data precision to protect privacy while maintaining analytical utility.
The Core Idea
When achieving k-anonymity (making each record indistinguishable from at least k-1 others), you need practical methods to hide identifying details. Generalization and suppression are your two main tools for reducing the precision of quasi-identifiers.
Generalization means replacing specific values with broader categories. Instead of an exact age like "42," you might report an age range like "40-49." Instead of a precise ZIP code like "94102," you generalize to "941**" (covering multiple neighborhoods). Think of it like zooming out a map—details blur together, making individuals harder to pinpoint while keeping the data useful for analysis.
Suppression means completely removing or masking certain values—typically outliers or combinations that would otherwise require excessive generalization. If only one person in your dataset is 102 years old, that extreme value could identify them. Rather than generalizing everyone's age to "0-120" (destroying utility), you suppress that single outlier by replacing it with a placeholder like "*" or simply omitting that field.
Choosing Between Techniques
Use generalization for the bulk of your dataset—it preserves patterns and allows aggregate analysis. Use suppression sparingly for exceptional cases that resist anonymization, accepting some data loss to protect outliers without destroying everyone's data quality.
The art lies in balancing privacy protection with data utility: over-generalize and your data becomes meaningless; under-generalize and re-identification risks remain.
Key Takeaway: Generalization reduces precision by replacing specific values with broader ranges or categories, while suppression removes problematic values entirely—together they're the practical techniques for implementing k-anonymity in real datasets.