Data Anonymization

A data protection technique where direct identifiers in a dataset are replaced with artificial identifiers or pseudonyms. Unlike anonymization, this process is reversible with the use of a separate, securely stored key. It is a core requirement under regulations like the GDPR, which treats pseudonymized data as personal data, still subject to protection. Common techniques include tokenization and hashing with a salt.

• Key Difference from Anonymization: Pseudonymization reduces, but does not eliminate, the linkability to a data subject.
• Use Case: Often used in analytics and testing environments where data utility must be preserved for internal processes while mitigating direct identification risk.

A rigorous mathematical framework for quantifying and limiting the privacy loss incurred when an individual's data is included in a statistical analysis. It provides a provable privacy guarantee by adding carefully calibrated statistical noise to query results or to the dataset itself. A common implementation is the epsilon (ε)-differential privacy model, where a smaller ε value indicates stronger privacy.

• Core Principle: The output of an analysis should be roughly the same whether any single individual's data is included or excluded.
• Application: Used by organizations like Apple and the U.S. Census Bureau to collect aggregate insights without exposing individual records.

A formal privacy model that requires each record in a published dataset to be indistinguishable from at least k-1 other records with respect to a set of quasi-identifiers (e.g., ZIP code, birth date, gender). This is achieved through techniques like generalization (e.g., replacing a specific age with an age range) and suppression (removing outlier records).

• Limitation: While it protects against identity disclosure, it is vulnerable to homogeneity attacks (if all k records share the same sensitive attribute) and background knowledge attacks.
• Foundation: Serves as the basis for more advanced models like l-diversity and t-closeness.

The process of creating artificial datasets that mimic the statistical properties and relationships of a real dataset without containing any actual personal information. Models like Generative Adversarial Networks (GANs) or variational autoencoders learn the underlying data distribution to produce new, synthetic records.

• Primary Use: Bypasses data scarcity, preserves privacy for model training, and creates data for testing edge cases.
• Privacy Consideration: The generative model itself must be trained securely, often using differentially private training to prevent memorization of real data points.

A broad set of techniques for obfuscating specific data within a dataset to protect confidential information. It is often applied to non-production environments like development or testing. Methods include:

• Static Data Masking: Irreversibly transforming data in a copy of the database.
• Dynamic Data Masking: Applying masking rules in real-time based on user roles, leaving the source data unchanged.
• Techniques: Substitution (swapping values), shuffling, encryption, and nulling out.

Key Distinction: While similar to anonymization, masking is often a reversible process (e.g., using encryption) and may be applied selectively to specific fields rather than aiming for full de-identification.

A form of encryption that allows specific types of computations to be performed directly on ciphertext, generating an encrypted result that, when decrypted, matches the result of the operations as if they had been performed on the plaintext. This enables privacy-preserving computation where data can be processed by an untrusted party (e.g., a cloud server) without ever being decrypted.

• Categories: Partially Homomorphic Encryption (PHE) supports one operation (addition or multiplication), Somewhat Homomorphic Encryption (SHE) supports limited operations, and Fully Homomorphic Encryption (FHE) supports arbitrary computations.
• Application in Anonymization: While not anonymization per se, it is a foundational technology for secure multi-party computation and federated learning, enabling analytics on encrypted, sensitive data.