ppdm in social network 1st part

15
Privacy preserving data mining in social network سیده حمیده رسولی21 مهر94

Upload: hamide-rasouli

Post on 11-Apr-2017

236 views

Category:

Data & Analytics


0 download

TRANSCRIPT

Page 1: ppdm in social network 1st part

Privacy preserving data mining in social network

رسولی حمیده سیده

94 مهر 21

Page 2: ppdm in social network 1st part

Social Network

•Graph structure

•Entities & Connections •Private information

2

Page 3: ppdm in social network 1st part

Data Mining in Social Network

•Legitimate

•Illegitimate inference attacks

Page 4: ppdm in social network 1st part

Privacy Concerns of Individuals in Social Network

•Privacy after release

anonymization

naive anonymity quasi Identifiers

K-anonymity

Random-based noise addition

•Private information leakage

inference

On-line & Released

4

Page 5: ppdm in social network 1st part

Chaleneges

•From data mining perspective

high-dimention large-scale

•From privacy preserving perspective Confidential information is unknown

Mathematically defining data is hard

5

Page 6: ppdm in social network 1st part

Machine learning techniques to be discussed

Clustering

Naive bayes

6

Page 7: ppdm in social network 1st part

CLUSTERING

•Unsupervised

•Applications •Goal •A good clustering

•Similarity factor

Page 8: ppdm in social network 1st part

1-Partitioning

K-means cost=ktn

CLUSTERING APPROACHES

8

Page 9: ppdm in social network 1st part

CLUSTERING APPROACHES

•2-Hierarchical Divisive(Top-Down) Agglomerative(Down-Top)

9

Page 10: ppdm in social network 1st part

Bayesian Learning •Bayes theorem hi member of H

Maximum A Posteriori (MAP) hypothesis

hmap= max{(P(h1|D), P(h2|D),……} P(AB)=p(A|B).p(B)=p(B|A).p(A)

H= hypothesis D= data

Page 11: ppdm in social network 1st part

•Naïve Bayes classifier F:X V , X: (a1,a2,a3,…,an)

P()=p(a1,a2,….).p(a1).p(a2)…. Naïve bayes assumption

Bayesian Learning

11

Page 12: ppdm in social network 1st part

Resources-1

•1-https://en.wikipedia.org/wiki/Naive_Bayes_classifier • •2-http://ceit.aut.ac.ir/~shiry/lecture/machine learning/tutorial/clustering. • •3-T.Tassa and J.Cohen,"Anonymization of Centralized and Distributed Social Networks by Sequential Clustering "IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING, VOL. 25, NO. 2,FEBRUARY 2013. • •4-R.Heatherly, M.Kantarcioglu,and B.Thuraisingham,"Preventing Private Information Inference Attacks on Social Networks Technical Report UTDCS-03-09" ,Computer Science Department University of Texas at Dallas,2009. • •5-Raymond D. Heatherly,"PRIVACY-PRESERVING SOCIAL NETWORK ANALYSIS",The University of Texas at Dallas, 2011. •

12

Page 13: ppdm in social network 1st part

Resources-2 •6-Liu, Lian, "PRIVACY PRESERVING DATA MINING FOR NUMERICAL MATRICES, SOCIAL NETWORKS, AND BIG DATA" (2015). Theses and Dissertations--Computer Science. Paper 31. • •7-Benjamin C. M. Fung, Yan’an Jin,"Preserving Privacy and Frequent Sharing Patterns for Social Network Data Publishing",IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining, 2013. • •8-X. Luo, J.X. Yu, and Z. Li (Eds.): ADMA 2014, LNAI 8933, pp. 267–278, 2014. • •9-B. Thompson,D. Yao,"The Union-Split Algorithm and Cluster-Based Anonymization of Social Networks",ASIACCS ’09, March 10-12, 2009, Sydney, NSW, Australia. • •10-B.calco,Shamsuddin S. Khan,"Privacy Preserving Data Mining for Social Networks",978-1-4799-7319-4/14/$31.00 ©2014 IEEE.

13

Page 14: ppdm in social network 1st part

Resources-3

•-11 كیوان‌پور محمدرضا کاوی داده در پیشرفته مباحث - حسن‌زاده فرانك - مرادی محمد . •-12 ماشین یادگیري Tom M. Mitchell ي نوشته زعیم نخبه محمد ي ترجمه

14

Page 15: ppdm in social network 1st part

thanks