Anonymization of Multidimensional Data
• Introduction
• Classification of Privacy Preserving Methods
• Classification of Data in Multi-dimensional
Dataset
• Group Based Anonymization Technique
– k-Anonymity
– l-Diversity
– t-Closeness
Program Name: Program Code:
Introduction
• Static Anonymization
• Multi-Dimensional or Relational Data
• Easy way to represent Data
• Easy target for Privacy Attacks
Program Name: Program Code:
Classification of Privacy Preserving Methods
Program Name: Program Code:
Classification of Data in Multi-dimensional Dataset
• Protecting Explicit Identifier
• Protecting Quasi Identifier
• Protecting Sensitive Data
Program Name: Program Code:
Protecting Explicit Identifier
• A Multi dimensional Dataset consists of attributes of different data
types, such as numeric, Strings and so on.
• Data anonymization methods should focus on semantics of the data
and not the syntax.
• Understand the semantics of the data in the context of application,
So as to apply correct/appropriate anonymization technique on the
data.
Program Name: Program Code:
Protecting Explicit Identifier
Program Name: Program Code:
Protecting Explicit Identifier
• EI identify the record owner explicitly, and as such should be masked
completely.
• While masking these fields, it is critical to take care of two aspects.
– Referential Integrity
– Consistently masking across databases.
Program Name: Program Code:
Protecting Explicit Identifier
Program Name: Program Code:
Protecting Explicit Identifier
Program Name: Program Code:
Protecting Explicit Identifier
Program Name: Program Code:
Protecting Explicit Identifier
Program Name: Program Code:
Protecting Explicit Identifier
Substitution Mechanism
Program Name: Program Code:
Protecting Explicit Identifier Substitution Mechanism
Program Name: Program Code:
Protecting Quasi-Identifier
• Masking EI alone is not sufficient, as an adversary
can still use QI to re-identify the record owner.
• This linking is called record linkage where a record
the from the database is linked with a record in a
external data source.
• Hospital Record
ID F_Name L_Name Gender Address DOB ZIP Disease
12432 M MA 21/02/1946 01880 Cancer
Program Name: Program Code:
Protecting Quasi-Identifier
• Voters List
Voter ID F_Name L_Name Gender Address DOB ZIP
893423 Weld M MA 21/02/1946 01880
• There are two aspects that need to be considered while anonymizing
QI:
– The analytical utility of the QI needs to be preserved.
– the correlation QI attributes with sensitive data needs to be maintained to
support the utility of anonymized data.
Program Name: Program Code:
Protecting Quasi-Identifier
• Challenges in Protecting QI:
– Identifying the boundary between QI and SD and
anonymizing the QI is probably the toughest problem to
solve in Privacy Preservation.
– The main challenges in anonymizing QI attributes are:
• High Dimensionality
• Background Knowledge of the adversary
• Availability of External Knowledge
• Correlation with SD to ensure Utility
• Maintaining Analytical Utility
Program Name: Program Code:
Protecting Quasi-Identifier
• To Protecting QI, we need to use group based
anonymization techniques:
– k-Anonymity
– l-Diversity
– t-closeness
Program Name: Program Code:
Protecting Sensitive Data
• If sensitive data in original form, then it provides a
channel for re-identification
Program Name: Program Code:
Protecting Sensitive Data
Program Name: Program Code: