Capability Access Control for P2P Data Sharing - PowerPoint PPT Presentation

1 / 32
About This Presentation
Title:

Capability Access Control for P2P Data Sharing

Description:

P2P: BitTorrent, Kazaa. Web services: Flickr, YouTube. Integrate these two functions? ... P2P: Kazaa, BitTorrent. Access control. ACLs. Cryptography. Capabilities ... – PowerPoint PPT presentation

Number of Views:99
Avg rating:3.0/5.0
Slides: 33
Provided by: csWash
Category:

less

Transcript and Presenter's Notes

Title: Capability Access Control for P2P Data Sharing


1
Capability Access Control for P2P Data Sharing
  • Roxana Geambasu
  • Magda Balazinska
  • Steve Gribble
  • Hank Levy

2
Motivation
  • Lots of personal data
  • Data organization
  • Desktop search, virtual folders Spotlight
  • Sharing
  • P2P BitTorrent, Kazaa
  • Web services Flickr, YouTube
  • Integrate these two functions?

3
Motivating Scenario
Alice
Bob
Main Dishes
Snacks
Grandpa
European
Asian
Alice's recipes
Recipes
Recipes
sharing
view composition
view
4
SharedViews
  • Organize data into views
  • Selectively share dynamic collections
  • Integrate other dynamic collections locally
  • Protect shared data
  • Peer-to-peer
  • No global administration, less management
  • Assumption read-only workloads

5
Agenda
  • Motivation
  • Approach
  • Views and capabilities
  • Integration of capabilities with views
  • The SharedViews system
  • Capability implementation
  • Query processing
  • Prototype
  • Evaluation

6
Approach Views and Capabilities
  • Integrate DB views with capability protection
  • Views
  • Organize data
  • Share dynamic sets of data
  • Seamlessly compose views
  • Capabilities
  • Control access to views
  • Facilitate sharing exchanging capabilities
  • Ease management

7
Scenario SharedViews Solution
Alice
Bob
CA2
Main Dishes
Snacks
Grandpa
CG1
European
Asian
CG1
CG0
CA0
Grandpa's recipes
Alice's recipes
Files
Files
CX
CX
CX
CX
Capa to a view
View defined on a capa
Capa sharing
8
Integration of Views and Capabilities
  • Views are named via capabilities
  • Capabilities enable same rights as DB
  • Other rights CATALOG_LOOKUP
  • Integrate capabilities into SQL
  • Minor modifications

9
Query Language Modifications
  • Use capabilities to name views
  • Queries on top of capabilities
  • SELECT FROM CA1
  • WHERE CONTAINS(text,ginger)
  • CREATE VIEW returns capa
  • CREATE VIEW AS
  • SELECT FROM CG1 WHERE CONTAINS(text,snack)
  • UNION SELECT FROM CA0 WHERE CONTAINS(text,
    snack)
  • CA1

A

10
Query Language Modifications (2)
  • Create restricted capabilities
  • RESTRICT CA1 RIGHTS SELECT CA1
  • Usage selective sharing/revocation
  • Revoke capabilities
  • REVOKE CA1 USING CA1
  • CA1 requires REVOKE right
  • Other modifications
  • Bootstrapping
  • Catalog information lookup

A
B
G

11
Agenda
  • Motivation
  • Approach
  • Capabilities
  • Integration of capabilities with views
  • The SharedViews system
  • Capability structure
  • Query processing
  • Prototype
  • Evaluation

12
SharedViews Architecture
13
Capability Structure
  • Password capabilities
  • Probabilistic protection against forgery
  • Capability structure
  • Advantages of our model
  • No special privilege required to manage capas
  • Can be easily transferred

14
Query Processing
  • Validate capa at every invocation
  • Two techniques
  • Recursive evaluation
  • Query rewrite

15
Recursive Evaluation
A
B
G

16
Query Rewrite
A
B
G

17
The SharedViews Prototype
  • Query engine Beagle
  • GUI
  • Web browser Web server
  • People fill in Web forms to operate on views
  • Capabilities are exposed as Web links
  • Capabilities are bookmarked
  • Share a view by emailing the Web link (capa)
  • Successful session is similar to search session

18
Agenda
  • Motivation
  • Approach
  • Capabilities
  • Integration of capabilities with views
  • The SharedViews system
  • Capability implementation
  • Query processing
  • Prototype
  • Evaluation

19
Evaluation
  • Dominating components?
  • Scales with query distribution?
  • Opportunities to optimize?

20
Dominating Components Simple queries
21
Scaling with Query Distribution
  • Query distribution
  • Depth
  • Breadth
  • Fast networks
  • Insignificant overhead for recursive evaluation
  • Slow networks
  • Recursive return of results becomes bottleneck

22
Query Rewrite and Optimization
23
Summary
  • SharedViews
  • Selective and protected sharing
  • No global protection structure
  • Seamlessly integrate remote data collections into
    local one
  • A web of personal files
  • Use views to organize and share
  • Use capabilities to protect

24
Appendix
25
Motivation
  • Huge collections of personal data
  • Data organization
  • Desktop search, virtual folders Spotlight
  • Sharing
  • P2P BitTorrent, Kazaa
  • Web services Flickr, YouTube
  • Integrate these two functionalities?

26
Capabilities
  • Capability secure token that
  • identifies an object
  • enables a set of rights
  • provides holder with authority
  • must be unforgeable
  • Facilitate sharing
  • Ease management
  • No user accounts
  • No global/coordinated protection

27
Capabilities Pros and Cons to ACLs
  • Ease sharing
  • Exchange capabilities
  • Simplify management
  • No user accounts
  • No user authentication
  • No centralized / global protection
  • But ACLs
  • Allow access logging and tracking
  • Allow confinement
  • Overall
  • Capabilities are suited in our unmanaged, p2p,
    home environment

28
Catalog Tables
  • Two catalog tables
  • CapTable -- stores protection-related info
  • All capabilities handed out by the system are
    saved
  • Selective capability revocation is easy

29
The SharedViews Prototype (2)
  • Successful session is similar to search session
  • User submits a view evaluation request
  • System returns Web link-like capas to files
  • User selects link to a file
  • System returns content a.o. attributes
  • Failures are handled, ensuring
  • Faulty view evaluation does not reveal unshared
    files
  • Best effort

30
Related Work
  • Data organization
  • Views WinFS, Spotlight, Google Desktop
  • Personal information management Haystack
  • Data sharing
  • P2P Kazaa, BitTorrent
  • Access control
  • ACLs
  • Cryptography
  • Capabilities
  • Hydra, The Capability-based System, Amoeba

31
Future Work
  • Scalability
  • Caching and replication
  • Real workload needed
  • Enable updates
  • Capability organization

32
Scaling with Query Distribution
  • Recursive evaluation overhead
  • Small for fast networks
  • Increased depth 1-5 3 increase for 5000
    results
  • Increase breadth 1-4 17 increase for 5000
    results
  • Large for broadband
  • Increased depth 1-5 80 increase for 5000
    results
  • Query rewrite vs recursive eval over broadband
  • View depth 5, and 5000 results 24 benefit
Write a Comment
User Comments (0)
About PowerShow.com