Data Anonymization - Generalization Algorithms Li Xiong CS573 Data Privacy and Anonymity.

Presentation on theme: "Data Anonymization - Generalization Algorithms Li Xiong CS573 Data Privacy and Anonymity."— Presentation transcript:

1 Data Anonymization - Generalization Algorithms Li Xiong CS573 Data Privacy and Anonymity

2 Generalization and Suppression Generalization Replace the value with a less specific but semantically consistent value #ZipAgeNationalityCondition 141076< 40*Heart Disease 248202< 40*Heart Disease 341076< 40*Cancer 448202< 40*Cancer Suppression Do not release a value at all Z0 = {41075, 41076, 41095, 41099} Z1 = {4107*. 4109*} Z2 = {410**} S0 = {Male, Female} S1 = {Person}

3 3 Complexity Search Space: Number of generalizations =  (Max level of generalization for attribute i + 1) attrib i If we allow generalization to a different level for each value of an attribute: Number of generalizations =  (Max level of generalization for attribute i + 1) attrib i #tuples

4 Hardness result Given some data set R and a QI Q, does R satisfy k-anonymity over Q? Easy to tell in polynomial time, NP! Finding an optimal anonymization is not easy NP-hard: reduction from k-dimensional perfect matching A polynomial solution implies P = NP A. Meyerson and R. Williams. On the complexity of optimal k-anonymity. In PODS’04.

5 Taxonomy of Generalization Algorithms Top-down specialization vs. bottom-up generalization Global (single dimensional) vs. local (multi- dimensional) Complete (optimal) vs. greedy (approximate) Hierarchy-based (user defined) vs. partition- based (automatic) K. LeFerve, D. J. DeWitt, and R. Ramakrishnan. Incognito: Efficient Full-Domain K-Anonymity. In SIGMOD 05

6 Generalization algorithms Early systems µ-Argus, Hundpool, 1996 - Global, bottom-up, greedy Datafly, Sweeney, 1997 - Global, bottom-up, greedy k-anonymity algorithms AllMin, Samarati, 2001 - Global, bottom-up, complete, impractical MinGen, Sweeney, 2002 - Global, bottom-up, complete, impractical Bottom-up generalization, Wang, 2004 – Global, bottom-up, greedy TDS (Top-Down Specialization), Fung, 2005 - Global, top-down, greedy K-OPTIMIZE, Bayardo, 2005 – Global, top-down, partition-based, complete Incognito, LeFevre, 2005 – Global, bottom-up, hierarchy-based, complete Mondrian, LeFevre, 2006 – Local, top-down, partition-based, greedy

7 Hundpool and Willenborg, 1996 Greedy approach Global generalization with tuple suppression Not guaranteeing k-anonymity µ-Argus

8 µ-Argus algorithm

9 µ-Argus

10 Problems With µ-Argus 1. Only 2- and 3- combinations are examined, there may exist 4 combinations that are unique – may not always satisfy k-anonymity 2. Enforce generalization at the attribute level (global) – may over generalize

11 The Datafly System Sweeney, 1997 Greedy approach Global generalization with tuple suppression

12 Datafly Algorithm Core Datafly Algorithm

13 Datafly MGT resulting from Datafly, k=2, QI={Race, Birthdate, Gender, ZIP}

14 Problems With Datafly 1.Generalizing all values associated with an attribute (global) 2.Suppressing all values within a tuple (global) 3.Selecting the attribute with the greatest number of distinct values as the one to generalize first – computationally efficient but may over generalize

15 Generalization algorithms Early systems µ-Argus, Hundpool, 1996 - Global, bottom-up, greedy Datafly, Sweeney, 1997 - Global, bottom-up, greedy k-anonymity algorithms AllMin, Samarati, 2001 - Global, bottom-up, complete, impractical MinGen, Sweeney, 2002 - Global, bottom-up, complete, impractical Bottom-up generalization, Wang, 2004 – Global, bottom-up, greedy TDS (Top-Down Specialization), Fung, 2005 - Global, top-down, greedy K-OPTIMIZE, Bayardo, 2005 – Global, top-down, partition-based, complete Incognito, LeFevre, 2005 – Global, bottom-up, hierarchy-based, complete Mondrian, LeFevre, 2006 – Local, top-down, partition-based, greedy

16 5/7/2015 16 K-OPTIMIZE Practical solution to guarantee optimality Main techniques Framing the problem into a set-enumeration search problem Tree-search strategy with cost-based pruning and dynamic search rearrangement Data management strategies

17 5/7/2015 17 Anonymization Strategies Local suppression Delete individual attribute values E.g. Global attribute generalization Replace specific values with more general ones for an attribute Numeric data: partitioning of the attribute domain into intervals. E.g. Age={[1-10],...,[91- 100]} Categorical data: generalization hierarchy supplied by users. E.g. Gender = [M or F]

18 5/7/2015 18 K-Anonymization with Suppression K-anonymization with suppression Global attribute generalization with local suppression of outlier tuples. Terminologies Dataset: D Anonymization: {a 1, …, a m } Equivalent classes: E v n,m v 1,n … v 1,m …v 1,1 a1a1 amam E{E{

19 5/7/2015 19 Finding Optimal Anonymization Optimal anonymization determined by a cost metric Cost metrics Discernibility metric: penalty for non- suppressed tuples and suppressed tuples Classification metric

20 5/7/2015 20 Modeling Anonymizations Assume a total order over the set of all attribute domain Set representation for anonymization E.g. Age:, Gender:, Marital Status: {1, 2, 4, 6, 7, 9} -> {2, 7, 9} Power set representation for entire anonymization space Power set of {2, 3, 5, 7, 8, 9} - order of 2 n ! {} – most general anonymization {2,3,5,7,8,9} – most specific anonymization

21 5/7/2015 21 Optimal Anonymization Problem Goal Find the best anonymization in the powerset with lowest cost Algorithm set enumeration search through tree expansion - size 2 n Top-down depth first search Heuristics Cost-based pruning Dynamic tree rearrangement Set enumeration tree over powerset of {1,2,3,4}

22 5/7/2015 22 Node Pruning through Cost Bounding Intuitive idea prune a node H if none of its descendents can be optimal Cost lower-bound of subtree of H Cost of suppressed tuples bounded by H Cost of non-suppressed tuples bounded by A H A

23 5/7/2015 23 Useless Value Pruning Intuitive idea Prune useless values that have no hope of improving cost Useless values Only split equivalence classes into suppressed equivalence classes (size < k)

24 5/7/2015 24 Tree Rearrangement Intuitive idea Dynamically reorder tree to increase pruning opportunities Heuristics sort the values based on the number of equivalence classes induced

25 5/7/2015 25 Experiments Adult census dataset 30k records and 9 attributes Fine: powerset of size 2 160 Evaluations of performance and optimal cost Comparison with greedy/stochastic method 2-phase greedy generalization/specialization Repeated process

26 5/7/2015 26 Results – Comparison None of the other optimal algorithms can handle the census data Greedy approaches, while executing quickly, produce highly sub- optimal anonymizations Comparison with 2-phase method (greedy + stochastic)

27 5/7/2015 27 Comments Interesting things to think about Domains without hierarchy or total order restrictions Other cost metrics Global generalization vs. local generalization

28 Generalization algorithms Early systems µ-Argus, Hundpool, 1996 - Global, bottom-up, greedy Datafly, Sweeney, 1997 - Global, bottom-up, greedy k-anonymity algorithms AllMin, Samarati, 2001 - Global, bottom-up, complete, impractical MinGen, Sweeney, 2002 - Global, bottom-up, complete, impractical Bottom-up generalization, Wang, 2004 – Global, bottom-up, greedy TDS (Top-Down Specialization), Fung, 2005 - Global, top-down, greedy K-OPTIMIZE, Bayardo, 2005 – Global, top-down, partition-based, complete Incognito, LeFevre, 2005 – Global, bottom-up, hierarchy-based, complete Mondrian, LeFevre, 2006 – Local, top-down, partition-based, greedy

29 Mondrian Top-down partitioning Greedy Local (multidimensional) – tuple/cell level

30 Global Recoding Mapping domains of quasi-identifiers to generalized or altered values using a single function Notation D x is the domain of attribute X i in table T Single Dimensional φ i : D xi  D’ for each attribute X i of the quasi- id φ i applied to values of X i in tuple of T

31 Local Recoding Multi-Dimensional Recode domain of value vectors from a set of quasi-identifier attributes φ : D x1 x … x D xn  D’ φ applied to vector of quasi-identifier attributes in each tuple in T

32 Partitioning Single Dimensional For each X i, define non-overlapping single dimensional intervals that covers D xi Use φ i to map x ε D x to a summary stat Strict Multi-Dimensional Define non-overlapping multi-dimensional intervals that covers D x1 … D xd Use φ to map (x x1 …x xd ) ε D x1 … D xd to a summary stat for its region

33 Global Recoding Example Multi-Dimensional Single Dimensional Partitions Age : {[25-28]} Sex: {Male, Female} Zip : {[53710-53711], 53712} Partitions {Age: [25-26],Sex: Male, Zip: 53711} {Age: [25-27],Sex: Female, Zip: 53712} {Age: [27-28],Sex: Male, Zip: [53710-53711]} k = 2 Quasi Identifiers Age, Sex, Zipcode

34 Global Recoding Example 2 k = 2 Quasi Identifiers Age, Zipcode Patient Data Single Dimensional Multi-Dimensional

35 Greedy Partitioning Algorithm Problem Need an algorithm to find multi-dimensional partitions Optimal k-anonymous strict multi-dimensional partitioning is NP-hard Solution Use a greedy algorithm Based on k-d trees Complexity O(nlogn)

36 Greedy Partitioning Algorithm

37 Algorithm Example k = 2 Dimension determined heuristically Quasi-identifiers Zipcode Age Patient Data Anonymized Data

38 Algorithm Example Iteration # 1 (full table) partition dim = Zipcode splitVal = 53711 ` LHSRHS fs

39 Algorithm Example continued Iteration # 2 (LHS from iteration # 1) partition dim = Age splitVal = 26 LHSRHS fs `

40 Algorithm Example continued Iteration # 3 (LHS from iteration # 2) partition No Allowable Cut ` Summary: Age = [25-26] Zip= [53711] Iteration # 4 (RHS from iteration # 2) partition No Allowable Cut ` Summary: Age = [27-28] Zip= [53710 - 53711] `

41 Algorithm Example continued Iteration # 5 (RHS from iteration # 1) partition No Allowable Cut ` Summary: Age = [25-27] Zip= [53712] `

42 Experiment Adult dataset Data quality metric (cost metric) Discernability Metric (C DM ) C DM = Σ EquivalentClasses E |E| 2 Assign a penalty to each tuple Normalized Avg. Eqiv. Class Size Metric (C AVG ) C AVG = (total_records/total_equiv_classes)/k

43 Comparison results Full-domain method: Incognito Single-dimensional method: K-OPTIMIZE

44 Data partitioning comparison

45 Mondrian