Efficient Hiding of Collaborative Recommendation Association Rules with Updates

1

Shyue-Liang Wang1, Ting-Zheng Lai3, Tzung-Pei Hong2, Yu-Lung Wu3 National University of Kaohsiung1,2

I-Shou University3,4

Kaohsiung, Taiwan

Efficient Hiding of Collaborative

Recommendation Association

2

Privacy Preserving Data Mining

• Privacy Preserving

– Given D_O – Minimum support – Minimum confidence – R_H • Rules to be hidden – Find D_M s.t. R_M R_O – R_H • D_M: modified database • R_M: modified rules

D

_O

R

_O

D

_M

_R

_M

DM

DM

Modification

3

Problem Description

₁

• Collaborative recommendation association rules

– R₃, R₅ same prediction as R₁to R₉

• Input: D_O, min_supp, min_conf, C

• Output: R_O TID Items T1 ABC T2 ABC T3 ABC T4 AB T5 A T6 AC min_supp=33% min_conf=70% 1 B=>A (66%, 100%) 2 C=>A (66%, 100%) 3 B=>C (50%, 75%) 4 C=>B (50%, 75%) 5 AB=>C (50%, 75%) 6 AC=>B (50%, 75%) 7 BC=>A (50%, 100%) 8 C=>AB (50%, 75%) 9 B=>AC (50%, 75%) 10 A=>B (66%, 66%) 11 A=>C (66%, 66%) 12 A=>BC (50%, 50%) D_O R_O: {R3, R5} |A|=6,|B|=4,|C|=4 |AB|=4,|AC|=4,|BC|=3 |ABC|=3 _Not AR

4

Problem Description

₂

• Input: D_O, H (items to be hidden

on RHS), min_supp, min_conf • Output: D_M, R_M TID Items T1 ABC T2 ABC T3 ABC T4 AB T5 A T6 AC min_supp=33% min_conf=70% H = {C}

AB

D_O, D_M R :{R1,R2,R7} 1 B=>A (66%, 100%) 2 C=>A (50%, 100%) 3 B=>C (50%, 50%) 4 C=>B (33%, 60%) 5 AB=>C (33%, 50%) 6 AC=>B (33%, 66%) 7 BC=>A (50%, 100%) 8 C=>AB (33%, 66%) 9 B=>AC (33%, 50%) 10 A=>B (66%, 66%) 11 A=>C (66%, 66%) 12 A=>BC (50%, 50%) |A|=6,|B|=4,|C|=3 |AB|=4,|AC|=3,|BC| =2,|ABC|=2 hidden hidden lost lost lost lost

5

Problem Description

₃

TID Items T1 ABC T2 ABC T3 ABC T4 AB T5 A T6 AC TID Items T7 BC T8 C T9 ABC TID Items T1 AB T2 ABC T3 ABC T4 AB T5 A T6 AC TID Items T7 B T8 C T9 ABC TID Items T1 ABC T2 ABC T3 ABC T4 AB T5 A T6 A T7 B T8 C T9 ABC (1)Combine, hide (DCBS) (2)Hide, combine(MSCR)

6

Numerical Experiments

₁

Time Effects Multiple Updates 0 100 200 300 400 500 10K 15K 20K 25K Data Size S ec o n d 1-Item-MSCR 1-Item-DCBS 2-Item-MSCR 2-Item-DCBS

7

Numerical Experiments

₂

Database Effects Database Effects 0% 2% 4% 6% 8% 10% 10k 15K 20K 25K Data Size P er ce n ta g e DCBS MSCR

8

Numerical Experiments

₃

Side Effects for MSCR

Multiple Updates Side Effects

0% 2% 4% 6% 8% 10K 15K 20K 25K Data Size P er ce n ta g e New Rules Lost Rules Hiding Failure

9

Numerical Experiments

₄

Side Effects for DCBS

DCBS Side Effects 0% 2% 4% 6% 8% 10K 15K 20K 25K Data Size P e rc e n ta g e New Rules Lost Rules Hiding Failure

10

Conclusion

• Study the problem of hiding collaborative

recommendation association rules with updates

• Automatically hide collaborative recommendation

association rule sets without pre-mining and

selection of a class of rules