mondrian
Class Mondrian

java.lang.Object
  anonymizer.Anonymizer
      mondrian.Mondrian

public class Mondrian
extends Anonymizer

Implementation of the Mondrian multi-dimensional partitioning algorithm for satisfying k-anonymity described in the following paper:

 @article{mondrian,
 author = {Kristen Lefevre and David J. Dewitt and Raghu Ramakrishnan},
 title = {Mondrian multidimensional k-anonymity},
 booktitle = {In ICDE},
 year = {2006}
 }
 

Our dimension selection heuristic is the one described in Section 4 of the paper. Namely, at each iteration, we "choose the dimension with the widest (normalized) range of values".

When multiple dimensions have the same width, our implementation chooses the first dimension that contains an allowable cut.

Given dimension, partitions are built by splitting the domain on the median value. If med represents the median, LHS contains all values <= med and RHS contains all values > med.

Field Summary

private java.lang.String[]

suppEq Generalized values for the suppression equivalence

Fields inherited from class anonymizer.Anonymizer

anonTable, conf, eqTable, sqlwrapper, suppressionThreshold

Constructor Summary

Mondrian(Configuration conf)
Class constructor

Method Summary

void	anonymize() Anonymizes the input.
void	changeKAnonymityRequirement(int newKvalue) Overwrites the privacy parameter k
protected AnonRecordTable	createAnonRecordsTable(java.lang.String tableName) Create anonRecords table
protected EquivalenceTable	createEquivalenceTable(java.lang.String tableName) Create equivalence table
java.lang.Double	getMedian(long eid, int att) Computes the median value on attribute indexed att for equivalence with ID eid
private double	getNormalizedWidth(java.lang.String gen, java.lang.String sup) Calculates the normalized width of a generalized value, based on the suppression value
protected void	insertTupleToAnonTable(java.lang.String[] vals, long eid) Insert a tuple to the anonRecords table
protected long	insertTupleToEquivalenceTable(java.lang.String[] vals) Insert a tuple to the equivalence table
private void	updateEIDs(long oldEID, long newEID, Interval range, int index) Updates the EIDs of equivalences with EID = oldEID to newEID if the predicate indicated by the interval is satisfied on the qi-attribute at the specified index

Methods inherited from class anonymizer.Anonymizer

anonymizeDataset, anonymizeDataset, isReadyForSuppression, main, outputResults_Anatomy, outputResults_GenVals, outputResults_GenValsDist, outputResults, readData, suppressEquivalences

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Detail

suppEq

private java.lang.String[] suppEq

Generalized values for the suppression equivalence

Constructor Detail

Mondrian

public Mondrian(Configuration conf)
         throws java.lang.Exception

Class constructor

Parameters:

conf - Configuration instance

Throws:

java.lang.Exception

Method Detail

changeKAnonymityRequirement

public void changeKAnonymityRequirement(int newKvalue)

Overwrites the privacy parameter k

Parameters:

newKvalue - New value of k

createAnonRecordsTable

protected AnonRecordTable createAnonRecordsTable(java.lang.String tableName)

Create anonRecords table

Specified by:

createAnonRecordsTable in class Anonymizer

Parameters:

tableName - Name of the table

Returns:

A new anonymization records table

createEquivalenceTable

protected EquivalenceTable createEquivalenceTable(java.lang.String tableName)

Create equivalence table

Specified by:

createEquivalenceTable in class Anonymizer

Parameters:

tableName - Name of the table

Returns:

A new equivalence table

insertTupleToEquivalenceTable

protected long insertTupleToEquivalenceTable(java.lang.String[] vals)
                                      throws java.lang.Exception

Insert a tuple to the equivalence table

Specified by:

insertTupleToEquivalenceTable in class Anonymizer

Parameters:

vals - All values of the tuple to be inserted (as read from the source, i.e., before any generalization)

Returns:

Equivalence id of the equivalence to which the tuple belongs

Throws:

java.lang.Exception

insertTupleToAnonTable

protected void insertTupleToAnonTable(java.lang.String[] vals,
                                      long eid)
                               throws java.lang.Exception

Insert a tuple to the anonRecords table

Specified by:

insertTupleToAnonTable in class Anonymizer

Parameters:

vals - All values of the tuple to be inserted (as read from the source, i.e., before any generalization)

eid - Equivalence id of the equivalence to which the tuple belongs

Throws:

java.lang.Exception

anonymize

public void anonymize()
               throws java.lang.Exception

Anonymizes the input. This function can be reused (with new k maybe), as long as eqTable and anonTable objects are fresh and the configuration has not changed.

Specified by:

anonymize in class Anonymizer

Throws:

java.lang.Exception

getNormalizedWidth

private double getNormalizedWidth(java.lang.String gen,
                                  java.lang.String sup)
                           throws java.lang.Exception

Calculates the normalized width of a generalized value, based on the suppression value

Parameters:

gen - String representation of the generalization interval

sup - String representation of the suppresion interval (i.e., the entire domain)

Returns:

normalized width of the generalization (genWidth / supWidth)

Throws:

java.lang.Exception

updateEIDs

private void updateEIDs(long oldEID,
                        long newEID,
                        Interval range,
                        int index)

Updates the EIDs of equivalences with EID = oldEID to newEID if the predicate indicated by the interval is satisfied on the qi-attribute at the specified index

Parameters:

oldEID - ID of the equivalence being split

newEID - New ID assigned to those records matching the range

range - The interval that specifies the new range on newEID for attribute at index

index - Index of the partitioning dimension

getMedian

public java.lang.Double getMedian(long eid,
                                  int att)
                           throws java.lang.Exception

Computes the median value on attribute indexed att for equivalence with ID eid

Parameters:

eid - Equivalence ID

att - An attribute index

Returns:

the median value

Throws:

java.lang.Exception