C3P: Context-Aware Crowdsourced Cloud Privacy Privacy Enhancing - - PowerPoint PPT Presentation

▶
c3p context aware crowdsourced cloud privacy
SMART_READER_LITE
LIVE PREVIEW

C3P: Context-Aware Crowdsourced Cloud Privacy Privacy Enhancing - - PowerPoint PPT Presentation

Sep 23, 2022 184 likes •968 views

C3P: Context-Aware Crowdsourced Cloud Privacy Privacy Enhancing Technologies Symposium, 2014 1 CloudSpaces Files to Flowers Conversion 2 Files to Flowers Conversion 2 Files to Flowers Conversion 2 Files to Flowers Conversion 2 Files

slide-1
SLIDE 1

C3P: Context-Aware Crowdsourced Cloud Privacy

1

CloudSpaces

Privacy Enhancing Technologies Symposium, 2014

slide-2
SLIDE 2

2

Files to Flowers Conversion

slide-3
SLIDE 3

2

Files to Flowers Conversion

slide-4
SLIDE 4

2

Files to Flowers Conversion

slide-5
SLIDE 5

2

Files to Flowers Conversion

slide-6
SLIDE 6

2

Files to Flowers Conversion

slide-7
SLIDE 7

3

60%

increase in corporate data shared to the cloud in 2015

Source: Elastica’s Q2 2015 Shadow Data Report

slide-8
SLIDE 8

3

20%

  • f files shared to the cloud contain protected data

60%

increase in corporate data shared to the cloud in 2015

Source: Elastica’s Q2 2015 Shadow Data Report

slide-9
SLIDE 9

3

20%

  • f files shared to the cloud contain protected data

60%

  • f sensitive files contain PII

30%

…contain health info

60%

increase in corporate data shared to the cloud in 2015

Source: Elastica’s Q2 2015 Shadow Data Report

slide-10
SLIDE 10

3

20%

  • f files shared to the cloud contain protected data

60%

  • f sensitive files contain PII

30%

…contain health info

Emergence of “Shadow IT”

60%

increase in corporate data shared to the cloud in 2015

Source: Elastica’s Q2 2015 Shadow Data Report

slide-11
SLIDE 11

You cannot use cloud services. You are fully protected. Your files are always encrypted before uploading.

Anti-Snooping Tools for the Cloud

Examples:

4

slide-12
SLIDE 12

You cannot run software. You are fully protected. Your files are always quarantined.

What if Antivirus Software was Similar?

5

slide-13
SLIDE 13

Obstacles

Privacy vs. Services dilemma

slide-14
SLIDE 14

Obstacles

Privacy vs. Services dilemma Context-dependence

  • f privacy
slide-15
SLIDE 15

Obstacles

I dedicate the rest of my life for sorting out sensitive from non-sensitive files on my HD

Privacy vs. Services dilemma Context-dependence

  • f privacy

Manual effort and expertise for assessing data sensitivity

6

slide-16
SLIDE 16

What is needed?

Ensure serviceable protection instead of brute encryption.

slide-17
SLIDE 17

What is needed?

Ensure serviceable protection instead of brute encryption. Account for the metadata, sharing environment, and data content.

slide-18
SLIDE 18

What is needed?

I dedicate the rest of my life for sorting out sensitive from non-sensitive files on my HD

Ensure serviceable protection instead of brute encryption. Account for the metadata, sharing environment, and data content. Automatically estimate the sensitivity of shared data.

7

slide-19
SLIDE 19

Introducing C3P

Various levels of information hiding

8

slide-20
SLIDE 20

Introducing C3P

Define data in terms of context Various levels of information hiding

8

slide-21
SLIDE 21

Introducing C3P

I dedicate the rest of my life for sorting out sensitive from non-sensitive files on my HD

Private crowdsourcing mechanism for gathering people privacy policies Define data in terms of context Various levels of information hiding

8

slide-22
SLIDE 22

Introducing C3P

I dedicate the rest of my life for sorting out sensitive from non-sensitive files on my HD

Private crowdsourcing mechanism for gathering people privacy policies Psychologically grounded approach for estimating sensitivity Define data in terms of context Various levels of information hiding

8

slide-23
SLIDE 23

Fine-Grained Policies

9

slide-24
SLIDE 24

Defining Data through Context

10

Content Metadata Environment

slide-25
SLIDE 25

Defining Data through Context

10

Content Metadata Environment

slide-26
SLIDE 26

Defining Data through Context

10

Content Metadata Environment

slide-27
SLIDE 27

Location Data Topic Media Home Office Document Software Financial Educational

Context V

  • cabulary

11

slide-28
SLIDE 28

Privacy Preserving Crowdsourcing

12

Business Me Colleague Financial Me Stranger

Faces Home Friend Financial Me Stranger Business Me Colleague Faces Home Friend

I dedicate the rest of my life for sorting out sensitive from non-sensitive files on my HD

User 1 User 2 User 3

slide-29
SLIDE 29

Privacy Preserving Crowdsourcing

12

Business Me Colleague Financial Me Stranger

Faces Home Friend Financial Me Stranger Business Me Colleague Faces Home Friend Faces Home Friend

Sharing Operation Context

I dedicate the rest of my life for sorting out sensitive from non-sensitive files on my HD

User 1 User 2 User 3

slide-30
SLIDE 30

Privacy Preserving Crowdsourcing

12

Business Me Colleague Financial Me Stranger

Faces Home Friend Financial Me Stranger Business Me Colleague Faces Home Friend Faces Home Friend

Work Sea Colleague Family Sharing Operation Context

I dedicate the rest of my life for sorting out sensitive from non-sensitive files on my HD

User 1 User 2 User 3

slide-31
SLIDE 31

Privacy Preserving Crowdsourcing

12

Business Me Colleague Financial Me Stranger

Faces Home Friend Financial Me Stranger Business Me Colleague Faces Home Friend Faces Home Friend

Work Sea Colleague Family Forward-Anonymity K-anonymity Sharing Operation Context

I dedicate the rest of my life for sorting out sensitive from non-sensitive files on my HD

User 1 User 2 User 3

Faces Home Friend

Context

slide-32
SLIDE 32

Sensitivity Estimation using Item Response Theory

13

Faces Home Friend

slide-33
SLIDE 33

Sensitivity Estimation using Item Response Theory

13

Faces Home Friend

High Sensitivity 75%

slide-34
SLIDE 34

Sensitivity Estimation using Item Response Theory

13

Faces Home Friend

High Sensitivity 75%

slide-35
SLIDE 35

High Privacy Attitude 75%

Sensitivity Estimation using Item Response Theory

13

Faces Home Friend

High Sensitivity 75%

slide-36
SLIDE 36

High Privacy Attitude 75%

Sensitivity Estimation using Item Response Theory

13

Faces Home Friend

High Sensitivity 75%

slide-37
SLIDE 37

High Privacy Attitude 75%

Sensitivity Estimation using Item Response Theory

13

Faces Home Friend

High Sensitivity 75%

slide-38
SLIDE 38

High Privacy Attitude 75%

Sensitivity Estimation using Item Response Theory

13

Faces Home Friend

High Sensitivity 75% Group Invariance

Faces Home Friend Faces Home Friend

slide-39
SLIDE 39

High Privacy Attitude 75%

Sensitivity Estimation using Item Response Theory

13

Faces Home Friend

High Sensitivity 75% Group Invariance

Faces Home Friend Faces Home Friend

Item Invariance

slide-40
SLIDE 40

Connecting the Dots

14

Client Server

?

slide-41
SLIDE 41

Connecting the Dots

14

Client Server

?

slide-42
SLIDE 42

Connecting the Dots

14

Financial Me Stranger

Client Server

Context Extraction

?

slide-43
SLIDE 43

Connecting the Dots

14

Financial Me Stranger

Client Server

Context Extraction Sensitivity Request

?

slide-44
SLIDE 44

Connecting the Dots

14

Financial Me Stranger

Client Server

Sensitivity Reply

?

slide-45
SLIDE 45

Connecting the Dots

14

Financial Me Stranger

Client Server

Sensitivity Reply Policy Decision

?

slide-46
SLIDE 46

Connecting the Dots

14

Financial Me Stranger

Client Server

Policy Decision Data Sharing

?

slide-47
SLIDE 47

Connecting the Dots

14

Financial Me Stranger

Client Server

Crowdsourcing

?

slide-48
SLIDE 48

Connecting the Dots

14

Financial Me Stranger

Client Server

Crowdsourcing

?

Sensitivity Computation

slide-49
SLIDE 49

Evaluation

15

C3P

slide-50
SLIDE 50

IRT Models Fit Privacy-Aware Cloud Sharing?

16

81 96

slide-51
SLIDE 51

IRT Models Fit Privacy-Aware Cloud Sharing?

  • Ex: With which privacy level would you

share a project presentation with a friend?

16

81 96

slide-52
SLIDE 52

IRT Models Fit Privacy-Aware Cloud Sharing?

  • Ex: With which privacy level would you

share a project presentation with a friend?

  • Standardized Infit Statistic:
  • (x-axis values should lie in [-2,2])

16

81 96

Dichotomous case Sensitivity Infit t-statistic A dot represents a context

slide-53
SLIDE 53

IRT Models Fit Privacy-Aware Cloud Sharing?

  • Ex: With which privacy level would you

share a project presentation with a friend?

  • Standardized Infit Statistic:
  • (x-axis values should lie in [-2,2])

16

81 96

Dichotomous case Sensitivity Infit t-statistic A dot represents a context

slide-54
SLIDE 54

IRT Models Fit Privacy-Aware Cloud Sharing?

  • Ex: With which privacy level would you

share a project presentation with a friend?

  • Standardized Infit Statistic:
  • (x-axis values should lie in [-2,2])

16

81 96

Polytomous case Infit t-statistic Sensitivity Dichotomous case Sensitivity Infit t-statistic A dot represents a context

slide-55
SLIDE 55

IRT Models Fit Privacy-Aware Cloud Sharing?

  • Ex: With which privacy level would you

share a project presentation with a friend?

  • Standardized Infit Statistic:
  • (x-axis values should lie in [-2,2])

16

81 96

Polytomous case Infit t-statistic Sensitivity Dichotomous case Sensitivity Infit t-statistic A dot represents a context

slide-56
SLIDE 56

IRT Models Fit Privacy-Aware Cloud Sharing?

  • Ex: With which privacy level would you

share a project presentation with a friend?

  • Standardized Infit Statistic:
  • (x-axis values should lie in [-2,2])

16

81 96

Yes!

Polytomous case Infit t-statistic Sensitivity Dichotomous case Sensitivity Infit t-statistic A dot represents a context

slide-57
SLIDE 57

Temporal Cost of Crowdsourcing & Privacy

17

Zipf context distribution

500 3125 30000

av.: 1 Item/6 hours

  • Synthetic Dataset:
slide-58
SLIDE 58

Temporal Cost of Crowdsourcing & Privacy

k

17

Zipf context distribution

500 3125 30000

av.: 1 Item/6 hours

  • Synthetic Dataset:
slide-59
SLIDE 59

Temporal Cost of Crowdsourcing & Privacy

k

17

Zipf context distribution

500 3125 30000

av.: 1 Item/6 hours

  • Synthetic Dataset:

Crowdsourcing cost: Hit rate (HR) from 0 to 90% in 10 days

slide-60
SLIDE 60

Temporal Cost of Crowdsourcing & Privacy

k

17

Zipf context distribution

500 3125 30000

av.: 1 Item/6 hours

  • Synthetic Dataset:

Crowdsourcing cost: Hit rate (HR) from 0 to 90% in 10 days Anonymity cost: HR difference starts high but vanishes in 10 days.

slide-61
SLIDE 61

Effect of Sharing Behavior on the Temporal Cost

18

Anonymity Parameter K=3

500 3125 30000

av.: 1 Item/6 hours

  • Synthetic Dataset:
slide-62
SLIDE 62

Effect of Sharing Behavior on the Temporal Cost

18

Effect: Long tail distribution is of lower temporal cost.

Anonymity Parameter K=3

500 3125 30000

av.: 1 Item/6 hours

  • Synthetic Dataset:
slide-63
SLIDE 63

Robustness Towards Malicious Users?

19

  • Test:
  • Assign sensitivities to items

and attitudes to people.

  • Honest users choose policies

according to the model.

  • Malicious users choose

policies at random.

slide-64
SLIDE 64

Robustness Towards Malicious Users?

19

  • Test:
  • Assign sensitivities to items

and attitudes to people.

  • Honest users choose policies

according to the model.

  • Malicious users choose

policies at random.

slide-65
SLIDE 65

Robustness Towards Malicious Users?

19

  • Test:
  • Assign sensitivities to items

and attitudes to people.

  • Honest users choose policies

according to the model.

  • Malicious users choose

policies at random.

slide-66
SLIDE 66

Robustness Towards Malicious Users?

19

  • Test:
  • Assign sensitivities to items

and attitudes to people.

  • Honest users choose policies

according to the model.

  • Malicious users choose

policies at random.

Preset Sensitivity Computed Sensitivity

  • Check
slide-67
SLIDE 67

Robustness Towards Malicious Users?

19

  • Test:
  • Assign sensitivities to items

and attitudes to people.

  • Honest users choose policies

according to the model.

  • Malicious users choose

policies at random.

Preset Sensitivity Computed Sensitivity

  • Check

Tolerance: 25% malicious: ≈8% difgerence, 50% malicious: ≈17% difgerence

slide-68
SLIDE 68

Future Work

  • Recommendation of policies to users
  • Batch sensitivity analysis
  • Considering probabilistic attacks on the scheme
  • Working with IRT alternatives.

20

ELO MF

slide-69
SLIDE 69

21

slide-70
SLIDE 70

22

PrivyShare

PrivyShare - Desktop

slide-71
SLIDE 71

PrivyShare Benefits

  • Works with any cloud service

23

slide-72
SLIDE 72

PrivyShare Benefits

  • Works with any cloud service
  • Provides “Sensitivity as a Service”

23

slide-73
SLIDE 73

PrivyShare Benefits

  • Works with any cloud service
  • Provides “Sensitivity as a Service”
  • Offers fine grained protection
  • Metadata cleaning
  • Face Blurring
  • Encryption
  • Encryption + Auxiliary Information (automatic summaries, blurred

thumbnails)

23

slide-74
SLIDE 74

24

PrivyShare

PrivyShare - Browser

slide-75
SLIDE 75

PrivySeal: Dealing with 3rd Party Cloud Apps

25

PrivySeal

privyseal.epfl.ch

slide-76
SLIDE 76

26

Questions

hamzaharkous.com

slide-77
SLIDE 77

Images/Media Credits

  • Pixel77
  • Freepik