beyond just data privacy bobji mungamuru hector garcia-molina christopher olston subhasish mitra
DESCRIPTION
CIDR 2007 Pacific Grove, CA, USA 3 Overview Configurations A neat way to capture these trade-offs Search machinery How to find good configurationsTRANSCRIPT
![Page 1: Beyond Just Data Privacy Bobji Mungamuru Hector Garcia-Molina Christopher Olston Subhasish Mitra](https://reader035.vdocument.in/reader035/viewer/2022062317/5a4d1b7b7f8b9ab0599b91aa/html5/thumbnails/1.jpg)
Beyond Just Data Privacy
Bobji MungamuruHector Garcia-Molina
Christopher OlstonSubhasish Mitra
![Page 2: Beyond Just Data Privacy Bobji Mungamuru Hector Garcia-Molina Christopher Olston Subhasish Mitra](https://reader035.vdocument.in/reader035/viewer/2022062317/5a4d1b7b7f8b9ab0599b91aa/html5/thumbnails/2.jpg)
CIDR 2007Pacific Grove, CA, USA
2
Competing Objectives
![Page 3: Beyond Just Data Privacy Bobji Mungamuru Hector Garcia-Molina Christopher Olston Subhasish Mitra](https://reader035.vdocument.in/reader035/viewer/2022062317/5a4d1b7b7f8b9ab0599b91aa/html5/thumbnails/3.jpg)
CIDR 2007Pacific Grove, CA, USA
3
Overview Configurations
A neat way to capture these trade-offs
Search machinery How to find good configurations
![Page 4: Beyond Just Data Privacy Bobji Mungamuru Hector Garcia-Molina Christopher Olston Subhasish Mitra](https://reader035.vdocument.in/reader035/viewer/2022062317/5a4d1b7b7f8b9ab0599b91aa/html5/thumbnails/4.jpg)
CIDR 2007Pacific Grove, CA, USA
4
Configurations
S = splitC = copy
![Page 5: Beyond Just Data Privacy Bobji Mungamuru Hector Garcia-Molina Christopher Olston Subhasish Mitra](https://reader035.vdocument.in/reader035/viewer/2022062317/5a4d1b7b7f8b9ab0599b91aa/html5/thumbnails/5.jpg)
CIDR 2007Pacific Grove, CA, USA
5
Configurations
![Page 6: Beyond Just Data Privacy Bobji Mungamuru Hector Garcia-Molina Christopher Olston Subhasish Mitra](https://reader035.vdocument.in/reader035/viewer/2022062317/5a4d1b7b7f8b9ab0599b91aa/html5/thumbnails/6.jpg)
CIDR 2007Pacific Grove, CA, USA
6
Search Machinery
More Longevity
More Privacy
![Page 7: Beyond Just Data Privacy Bobji Mungamuru Hector Garcia-Molina Christopher Olston Subhasish Mitra](https://reader035.vdocument.in/reader035/viewer/2022062317/5a4d1b7b7f8b9ab0599b91aa/html5/thumbnails/7.jpg)
CIDR 2007Pacific Grove, CA, USA
7
Contributions Metrics to evaluate a configuration
Algorithm for searching for good configurations
Avoiding configurations that don't “make sense”
Future work – performance
![Page 8: Beyond Just Data Privacy Bobji Mungamuru Hector Garcia-Molina Christopher Olston Subhasish Mitra](https://reader035.vdocument.in/reader035/viewer/2022062317/5a4d1b7b7f8b9ab0599b91aa/html5/thumbnails/8.jpg)
CIDR 2007Pacific Grove, CA, USA
8
Summary
Google: bobji
![Page 9: Beyond Just Data Privacy Bobji Mungamuru Hector Garcia-Molina Christopher Olston Subhasish Mitra](https://reader035.vdocument.in/reader035/viewer/2022062317/5a4d1b7b7f8b9ab0599b91aa/html5/thumbnails/9.jpg)
CIDR 2007Pacific Grove, CA, USA
9
![Page 10: Beyond Just Data Privacy Bobji Mungamuru Hector Garcia-Molina Christopher Olston Subhasish Mitra](https://reader035.vdocument.in/reader035/viewer/2022062317/5a4d1b7b7f8b9ab0599b91aa/html5/thumbnails/10.jpg)
CIDR 2007Pacific Grove, CA, USA
10
So What? Summary: given N servers with known failure
characteristics, minimize data loss risk subject to an upper bound on break-in risk
Applications Measure effectiveness of existing systems Damage assessment Security breach probability functions
![Page 11: Beyond Just Data Privacy Bobji Mungamuru Hector Garcia-Molina Christopher Olston Subhasish Mitra](https://reader035.vdocument.in/reader035/viewer/2022062317/5a4d1b7b7f8b9ab0599b91aa/html5/thumbnails/11.jpg)
CIDR 2007Pacific Grove, CA, USA
11
Related Work Data preservation Data privacy Survivable storage @ CMU StorageSS @ NCSA Generalization: secret sharing
![Page 12: Beyond Just Data Privacy Bobji Mungamuru Hector Garcia-Molina Christopher Olston Subhasish Mitra](https://reader035.vdocument.in/reader035/viewer/2022062317/5a4d1b7b7f8b9ab0599b91aa/html5/thumbnails/12.jpg)
CIDR 2007Pacific Grove, CA, USA
12
Metrics Probability of
break-ins, P(Ө) Probability of
data loss, Q(Ө)
![Page 13: Beyond Just Data Privacy Bobji Mungamuru Hector Garcia-Molina Christopher Olston Subhasish Mitra](https://reader035.vdocument.in/reader035/viewer/2022062317/5a4d1b7b7f8b9ab0599b91aa/html5/thumbnails/13.jpg)
CIDR 2007Pacific Grove, CA, USA
13
Metrics Depth (3) Class (read-once) Terminals (4) Non-terminals (3) Allow groups (e.g., {a,b}) Deny groups (e.g., {c,d})
![Page 14: Beyond Just Data Privacy Bobji Mungamuru Hector Garcia-Molina Christopher Olston Subhasish Mitra](https://reader035.vdocument.in/reader035/viewer/2022062317/5a4d1b7b7f8b9ab0599b91aa/html5/thumbnails/14.jpg)
CIDR 2007Pacific Grove, CA, USA
14
Optimization
maximize longevity, given a lower bound on privacy
![Page 15: Beyond Just Data Privacy Bobji Mungamuru Hector Garcia-Molina Christopher Olston Subhasish Mitra](https://reader035.vdocument.in/reader035/viewer/2022062317/5a4d1b7b7f8b9ab0599b91aa/html5/thumbnails/15.jpg)
CIDR 2007Pacific Grove, CA, USA
15
Optimization Configurations are isomorphic with the set of
factored monotone Boolean formulas
![Page 16: Beyond Just Data Privacy Bobji Mungamuru Hector Garcia-Molina Christopher Olston Subhasish Mitra](https://reader035.vdocument.in/reader035/viewer/2022062317/5a4d1b7b7f8b9ab0599b91aa/html5/thumbnails/16.jpg)
CIDR 2007Pacific Grove, CA, USA
16
Optimization Solution strategy:
Step 1: Find the best monotone DNF formula, F* Step 2: Find the best factorization of F*
Step 1 is finding a separating hyperplane in {0,1}n
Step 2 is well-studied in digital design literature