De#anonymizing,Social,Networks, and,Inferring,Private,Attributes, - - PowerPoint PPT Presentation

▶
de anonymizing social networks and inferring private
SMART_READER_LITE
LIVE PREVIEW

De#anonymizing,Social,Networks, and,Inferring,Private,Attributes, - - PowerPoint PPT Presentation

Aug 30, 2023 10 likes •352 views

De#anonymizing,Social,Networks, and,Inferring,Private,Attributes, Using,Knowledge,Graphs, Jianwei Qian Xiang#Yang Li Illinois Tech USTC,/Illinois Tech Chunhong Zhang Linlin Chen BUPT Illinois Tech Outline Background Prior Work Our Work

slide-1
SLIDE 1

De#anonymizing,Social,Networks, and,Inferring,Private,Attributes, Using,Knowledge,Graphs,

Jianwei Qian Xiang#Yang Li

Illinois Tech USTC,/Illinois Tech

Chunhong Zhang Linlin Chen

BUPT Illinois Tech

slide-2
SLIDE 2

Outline

De-anonymizing Social Networks and Inferring Private Attributes Using Knowledge Graphs

2

Background Prior Work Our Work Conclusion

slide-3
SLIDE 3

Background

  • Tons of social network data
  • Released to third-parties for research and business
  • Though user IDs removed, attackers with prior

knowledge can de-anonymize them. → privacy leak

De-anonymizing Social Networks and Inferring Private Attributes Using Knowledge Graphs

3

slide-4
SLIDE 4

Attacking Process

De-anonymizing Social Networks and Inferring Private Attributes Using Knowledge Graphs

4 Prior k.g.

slide-5
SLIDE 5

Prior k.g.

Privacy leaked!

De-anonymizing Social Networks and Inferring Private Attributes Using Knowledge Graphs

5

slide-6
SLIDE 6

Attack,Stage,1

De#Anonymization

De-anonymizing Social Networks and Inferring Private Attributes Using Knowledge Graphs

6

Which is Alice? Which is Bob?

Direct privacy leak

slide-7
SLIDE 7

Attack,Stage,2

Privacy Inference

  • Correlations between attributes/users

– Higher education => higher salary – Colleagues=> same company – Common hobbies => friends

  • Infer new info that is not published

De-anonymizing Social Networks and Inferring Private Attributes Using Knowledge Graphs

7

Indirect privacy leak

slide-8
SLIDE 8

What,Do,We,Want,to,Do? To understand

How privacy is leaked to the attacker

De-anonymizing Social Networks and Inferring Private Attributes Using Knowledge Graphs

8

slide-9
SLIDE 9

Outline

De-anonymizing Social Networks and Inferring Private Attributes Using Knowledge Graphs

9

Background Prior Work Our Work Conclusion

slide-10
SLIDE 10

Prior,Work

De-anonymizing Social Networks and Inferring Private Attributes Using Knowledge Graphs

10

  • Degree attack [SIGMOD’08]
  • 1-neighborhood attack [INFOCOM’13]
  • 1*-neighborhood attack [ICDE’08]
  • Friendship attack [KDD’11]
  • Community re-identification

[SDM’11]

  • k-degree anonymity
  • 1-neighborhood anonymity
  • 1*-neighborhood anonymity
  • "#-degree anonymity
  • k-structural diversity

De-anonymize one user

Fight

Never ending! Assume specific prior knowledge!

slide-11
SLIDE 11

Prior,Work

De-anonymize all the users

– Graph mapping based de-anonymization

[WWW’07, S&P’09, CCS’12, COSN’13, CCS’14, NDSS’15]

De-anonymizing Social Networks and Inferring Private Attributes Using Knowledge Graphs

11 Twitter Flickr Mapping

Attacker holds an auxiliary SN that

  • verlaps with the published SN
slide-12
SLIDE 12

Limitations

  • Assume attacker has specific prior knowledge

– We assume diverse and probabilistic knowledge

  • Focus on de-anonymization only. How attacker infers

privacy afterwards is barely discussed

– We consider it as 2nd attacking step!

De-anonymizing Social Networks and Inferring Private Attributes Using Knowledge Graphs

12

slide-13
SLIDE 13

Outline

De-anonymizing Social Networks and Inferring Private Attributes Using Knowledge Graphs

13

Background Prior Work Our Work Conclusion

slide-14
SLIDE 14

Goals

  • To construct a comprehensive and realistic model of

the attacker’s knowledge

  • To use this model to depict how privacy is leaked.

De-anonymizing Social Networks and Inferring Private Attributes Using Knowledge Graphs

14

slide-15
SLIDE 15

Challenges

  • Hard to build such an expressive model, given that

the attacker has various prior knowledge

  • Hard to simulate attacking process, since the

attacker has various techniques

De-anonymizing Social Networks and Inferring Private Attributes Using Knowledge Graphs

15

slide-16
SLIDE 16

Solution

Use knowledge graph to model attacker’s knowledge

De-anonymizing Social Networks and Inferring Private Attributes Using Knowledge Graphs

16

slide-17
SLIDE 17

Knowledge Graph

  • Knowledge => directed edge
  • Each edge has a confidence score

De-anonymizing Social Networks and Inferring Private Attributes Using Knowledge Graphs

17

slide-18
SLIDE 18

What’s Privacy?

  • Every edge is privacy
  • Privacy is leaked when $% e − $((*) > -(*)

De-anonymizing Social Networks and Inferring Private Attributes Using Knowledge Graphs

18 Prior Posterior

Say 30%

slide-19
SLIDE 19

De#Anonymization

argmax34567(8%,8:) 4567 8%,8: = ∑ 4(5, =)

(>,?)∈7

, S is node similarity function

De-anonymizing Social Networks and Inferring Private Attributes Using Knowledge Graphs

19

Prior knowledge 8% Anonymized graph 8:

Mapping A

slide-20
SLIDE 20

Node Similarity

  • Attribute Similarity

– Use Jaccard index to compare attribute sets

  • Relation similarity

– Inbound neighborhood – outbound neighborhood – l-hop neighborhood

4B 5, = = C>4> 5,= + CE4E 5, = + CF4F 5,=

4 5, = = CG4G 5, = + 1 − CG 4B 5, =

De-anonymizing Social Networks and Inferring Private Attributes Using Knowledge Graphs

20

slide-21
SLIDE 21

Problem Transformation

De-anonymizing Social Networks and Inferring Private Attributes Using Knowledge Graphs

21

8: 3I: (millions)

…… ……

Mapping => Max weighted bipartite matching I% 8%

Huge complexity!

Naïve3method:

slide-22
SLIDE 22

3I: (millions)

Top#k Strategy

De-anonymizing Social Networks and Inferring Private Attributes Using Knowledge Graphs

22

≤ "3I%

Suppose k=2

8:

3 1

I% 8%

Alice

slide-23
SLIDE 23

How,to,Choose, Top#k Candidates?

  • Intuition

– If two nodes match, their neighbors are also very likely to match.

  • Perform BFS on 8%

De-anonymizing Social Networks and Inferring Private Attributes Using Knowledge Graphs

23

Bob

2 1

Alice

slide-24
SLIDE 24

Complexity Analysis

Time Space

Building Bipartite Finding Matching

Naïve method I%I: [ I% + I: I%

#I:

[ I% + I:

#

Top-k strategy ≪ I%I: [ "#I%

]

[ "#I%

#

De-anonymizing Social Networks and Inferring Private Attributes Using Knowledge Graphs

24

Complexity3greatly reduced!

slide-25
SLIDE 25

Tradeoff

  • " balances accuracy and complexity
  • " = 10 is enough to achieve high accuracy

De-anonymizing Social Networks and Inferring Private Attributes Using Knowledge Graphs

25

0.2 0.4 0.6 0.8 1 10 20 30 40 50 60 70 80 k

sr=0.4 sr=0.6 sr=0.8

Accuracy

10 15 20 25 30 35 40 45 50 10 20 30 40 50 60 70 80 k

sr=0.4 sr=0.6 sr=0.8

Time

k k

slide-26
SLIDE 26

Privacy inference

De-anonymizing Social Networks and Inferring Private Attributes Using Knowledge Graphs

26

Predict new edges in knowledge graph

Kobe Bryant Nick Young teammate playInLeague teamInLeague

  • pponent

playFor LA Lakers playFor NY Knicks

slide-27
SLIDE 27

Path,Ranking Algorithm

  • Proposed by Ni Lao et al. in 2011 for a different topic
  • Correlations => “rules” => paths
  • Logistic regression

De-anonymizing Social Networks and Inferring Private Attributes Using Knowledge Graphs

27

Alice AIDS

slide-28
SLIDE 28

Experiments

  • Datasets

– Google+, Pokec

  • Steps

– Generate 8: – Generate 8% – Run the algorithms

De-anonymizing Social Networks and Inferring Private Attributes Using Knowledge Graphs

28

slide-29
SLIDE 29

De#Anonymization Results

Metrics: accuracy, run time

De-anonymizing Social Networks and Inferring Private Attributes Using Knowledge Graphs

29

10 20 30 40 50 60 70 80 Load Build Match Total Run Time(x102s)

RS EG RW

Accuracy 0.2 0.3 0.4 0.5 0.6 0.7 0.8 Accuracy

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.2 0.4 0.6 0.8 1 0 5 10 15 20 25 30 35 40 Accuracy Run Time(s) wA

Acc Time

0.2 0.4 0.6 0.8 1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Accuracy smin

k=5 k=10 k=15

De-anonymize about 60% of users

slide-30
SLIDE 30

Privacy Inference Results

De-anonymizing Social Networks and Inferring Private Attributes Using Knowledge Graphs

30

Metrics: hit@k, MRR (Mean reciprocal rank )

0.2 0.4 0.6 0.8 1 1.2 10 20 30 40 50 60 70 80 90 1 2 3 4 5 6 MRR(%) Hit@10(%) sr

MRR MRR,RG Hit Hit,RG

0.5 1 1.5 2 2.5 10 20 30 40 50 60 70 80 90 0 2 4 6 8 10 12 MRR(%) Hit@10(%) sr

MRR MRR,RG Hit Hit,RG

0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 5 10 15 20 25 30 35 MRR(%) Hit@10(%) Sample Ratio

MRR MRR,RG Hit Hit,RG

5 10 15 20 25 30 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 25 30 35 40 45 50 55 60 65 70 75 MRR(%) Hit@10(%) Sample Ratio

MRR MRR,RG Hit Hit,RG

Infers much more privacy than random guess

slide-31
SLIDE 31

Outline

De-anonymizing Social Networks and Inferring Private Attributes Using Knowledge Graphs

31

Background Prior Work Our Work Conclusion

slide-32
SLIDE 32

Conclusion

We have

  • Applied knowledge graphs to model the attacker’s

prior knowledge

  • Studied the attack process: de-anonymization &

privacy inference

  • Designed methods to perform attack
  • Done simulations and evaluations on two real

world social networks

De-anonymizing Social Networks and Inferring Private Attributes Using Knowledge Graphs

32

slide-33
SLIDE 33

Future work

  • Effective construction of the bipartite for large scale

social networks

  • Impact of adversarial knowledge on de-

anonymizability

  • Fine-grained privacy inference on the knowledge

graph

De-anonymizing Social Networks and Inferring Private Attributes Using Knowledge Graphs

33

slide-34
SLIDE 34

Thank you!

Jianwei Qian jqian15@hawk.iit.edu https://sites.google.com/site/jianweiqianshomepage

De-anonymizing Social Networks and Inferring Private Attributes Using Knowledge Graphs

34