Inferring Sibling Relationships on the Internet

1
Inferring Sibling Relationships on the Internet

• Anwar M. Haneef
• Santhosh R. Thampuran
• Department of Electrical and Computer Engineering
• University of Massachusetts, Amherst

2
Agenda

• Why Even Bother ? The Need
• Some Existing Algorithms
• SIBRE
• SIBRE And The Wider Picture
• SIBRE Heuristic
• Implementation
• Hurdles Faced
• Conclusions

3
Agenda

• Why Even Bother ? The Need
• Some Existing Algorithms
• SIBRE
• SIBRE And The Wider Picture
• SIBRE Heuristic
• Implementation
• Hurdles Faced
• Conclusions

4
Why Even Bother ? The Need

• Need to find relationships on the Internet
• Lack of centralized control
• De-centralized, uncontrolled Utopia getting
  harder to manage

5
Agenda

• Why Even Bother ? The Need
• Some Existing Algorithms
• SIBRE
• SIBRE And The Wider Picture
• SIBRE Heuristic
• Implementation
• Hurdles Faced
• Conclusions

6
Some Existing Algorithms

• The basic/refined autonomous system relationship
  inference algorithm mentioned during the course
• Results of the algorithm had poor sibling
  relationship identification. Could be because
• Error made by smaller in-experienced
  companies/ISPs
• Registries queried are not exhaustive

7
Agenda

• Why Even Bother ? The Need
• Some Existing Algorithms
• SIBRE
• SIBRE And The Wider Picture
• SIBRE Heuristic
• Implementation
• Hurdles Faced
• Conclusions

8
SIBRE Sibling Relationship

• Applies a heuristic based on the ASs location as
  recorded in the Internet Registry
• Querying the distributed Regional Internet
  Registries (Whois Servers)
• Computes using the information retrieved as to
  the geographic location of the AS whether there
  is a possibility of sibling relationships

9
Regional Internet Registries
Europe
Asia and Pacific
The Americas
.mil Top Level Domain
Provided by MERIT
10
Agenda

• Why Even Bother ? The Need
• Some Existing Algorithms
• SIBRE
• SIBRE And The Wider Picture
• SIBRE Heuristic
• Implementation
• Hurdles Faced
• Conclusions

11
SIBRE And The Wider Picture

• Where SIBRE fits in
• Basic/Refined algorithm mentioned earlier needs
  some means by which we can verify whether the
  sibling relationships identified may be true
• The information can be eye-balled, but an
  automated system was not available

12
Agenda

• Why Even Bother ? The Need
• Some Existing Algorithms
• SIBRE
• SIBRE And The Wider Picture
• SIBRE Heuristic
• Implementation
• Hurdles Faced
• Conclusions

13
The SIBRE Heuristic

• Background information we have
• AS to Country mapping (not exhaustive)
  information
• International Dialing Code to Country mapping
  information (from ATT)
• Distributed Regional Internet Registries

14
The SIBRE Heuristic

• AS to Country mapping (not exhaustive)
  information
• Initially check which country the given AS is
  located in
• International Dialing Code to Country mapping
  information (from ATT)
• This information may be used in case the country
  was not found in the previous database

15
International Dialing Code
3XX
7
4XX
1
8XX
9XX
2XX
5XX
6XX
16
The SIBRE Heuristic

• Regional Internet Registry
• This information is stored in RIPE 181/RPSL
  format (at least it should be ? )
• This contains more specific registration
  information related to the AS

17
(No Transcript)
18
Registry Identifier
19
Registry Identifier
20
Registry Identifier
SIBRE CORE
21
Registry Identifier
SIBRE CORE
22
Registry Identifier
SIBRE CORE
RESULTS
23
Agenda

• Why Even Bother ? The Need
• Some Existing Algorithms
• SIBRE
• SIBRE And The Wider Picture
• SIBRE Heuristic
• Implementation
• Hurdles Faced
• Conclusions

24
Implementation

• Using the whois UNIX command line interface for
  interacting with the Regional Internet Registry
• whois h ltwhoisservergt ltAsnumbergt
• Developed in Perl

25
Agenda

• Why Even Bother ? The Need
• Some Existing Algorithms
• SIBRE
• SIBRE And The Wider Picture
• SIBRE Heuristic
• Implementation
• Hurdles Faced
• Conclusions

26
Hurdles Faced

• Lack of standardization
• Some record maintainer details, some do not
• Some record country data, some do not
• Some record company name as the country, thank
  god some do not
• Give me a break !

27
Agenda

• Why Even Bother ? The Need
• Some Existing Algorithms
• SIBRE
• SIBRE And The Wider Picture
• SIBRE Heuristic
• Implementation
• Hurdles Faced
• Conclusions

28
Conclusions

• Ran over numerous test cases and compared with
  the output of the existing algorithm
• Geographically separated ISPs/organizations seen
  to form sibling relationships, which may not be
  accepted as valid using our heuristic
• SIBRE heuristic can not be used alone to
  determine sibling relationships, but may be used
  to verify results of other heuristics

29
Conclusions

• For example, German Department of Defense(DoD
  Education Activity Network Assistance Center)
  forms sibling relationship with US Department of
  Defense(Defense Logistics Agency).
• SPRINT (US), Etisalat (UAE), ATT(US), BBN(US)
  seen in one case to have a sibling to sibling
  relationship
• Worldwide Telecom(Japan), Verio (US), Teleglobe
  (Canada), Telephonica De ElSalvador (El Salvador)
  seem to have a sibling relationship.

30
References

• On Inferring Autonomous System Relationships in
  the Internet , L. Gao, IEEE Global Internet, Nov
  2000
• RFC 2622 Routing Policy Specification Language
• RFC 2650 Using RPSL in Practice
• RIPE-157 RIPE NCC Database Documentation

31
THANKS !!!