Reputation Network Analysis for Email Filtering

1
Reputation Network Analysis for Email Filtering

• Ravi Emani
• Ramesh Ravindran

2
Describes about

• E-mail Scoring mechanism based on a social
  network augmented with reputation ratings
• Algorithm for inferring reputation ratings
• Integration into a mail application TrustMail

3
Preventing Spam

• Trying to prevent spam from even reaching the
  users mailbox
• Methods
• - Whitelist filters
• - Social Networks
• - Connecting Users

4
Whitelist Filters

• Messages accepted according to a list of approved
  addresses created by the user
• Advantages
• - No spam in users inbox
• - Filters the spam into a low-priority
  folder
• Disadvantages
• -Extra burden on the user
• -Filters even the valid emails

5
Social Networks

• Proposed by Boykin and Roychowdhury
• Social network created from the messages received
  by the user
• Messages identified as spam, valid or unknown
  based on clustering thresholds and structural
  properties like the propensity for local
  clustering.
• Classifies about 50 of users email into spam or
  other valid categories

6
Optimization

• Extension of whitelisting and social network
  based filtering
• Uses a network that connects users
• A score of reputation or trust is assigned by
  the users to the people they know
• Results in a large reputation network connecting
  thousands of users
• Messages sorted by the score shown next to the
  messages in the inbox

7
Optimization

• Overcomes the problem of the whitelists
• More reliable than the whitelists even though the
  user takes the burden for creating an initial set
  of reputation ratings
• Less work comparatively

8
Creating the Reputation Network

• Uses a Distributed, web based social network
• Reputation rating inferred from one user to
  another
• Individuals are connected to each person they
  rated
• Results in a large interconnected network of users

9
How is it related to Semantic Web?

• The only requirement is that the individuals
  should assert their reputation ratings for one
  another in the network
• Individuals will be controlling their own data
• Data is maintained in a distributed fashion
• Data can be stored anywhere and integrated
  through a common foundation

10
Role of Semantic Web...

• Semantic web, along with its component languages
  RDF, RDFS, OWL utilize web architecture
• Supports distributed data management
• Users create ontologies with classes and
  properties and hence instances
• The instances of the classes help in describing
  the data on the web

11
FOAF Project

• Friend-Of-A-Friend project developed on Semantic
  Web
• An ontological vocabulary for describing people
  and their relationships
• Extended by providing a mechanism describing the
  reputation relationships
• Allows people to rate the reputation or
  trustworthiness of another person

12
Fig The reputation network developed as part of
the semantic web trust project at
http//trust.mindswap.org.
13
Algorithms for Inferring Reputation between
Individuals

• Recommendations are made to one person(source)
  about the reputation of another person(sink)
• Trust and reputation literature contains many
  different metrics
• These metrics are categorized according to the
  perspective used for making calculations

14
Perspective in Reputation Inference Algorithms

• Global metrics calculate a single value for each
  entity in the network
• Local metrics calculate a reputation rating for
  an individual in the network
• In global system an entity will always have the
  same inferred rating
• In local system an entity could be rated
  differently depending on the node the inference
  is made for

15
Perspective in Reputation Inference Algorithms

• Global metrics can be highly effective in
  situations where the experiences of users are
  similar
• Local metrics can be appropriate where users
  opinions vary about the same topic

E
9
1
D
C
10
10
A
B
16
Accurate Metrics for Inferring Reputation

• The inferred rating from the source to the sink
  is given by a weighted average of the neighbors
  reputation ratings of the sink.
• Reputation rating t from source i to sink s
  is written as tis
• No inference needed if source is directly
  connected to the sink
• If not, the reputation rating is calculated by
  weighted average of the reputation ratings
  returned for the sink by each of its n neighbors.

17

• getRating(source, sink)
• mark source as seen
• if source has no rating for sink
• denom 0
• num 0
• for each j in neighbors(source)
• if j has not been seen
• denom
• j2sink in(rating(source,j),ge
  tRating(j,sink))
• num rating(source,j) j2sink
• mark j unseen
• rating(source,sink) num/denom
• return rating(source,sink)

18
Accurate metrics for Inferring Reputation
The concise representation of how tis is weighted
is shown as follows
The condition in this formula ensures that the
source will never trust the sink more than any
intermediate node
19
Reputation Metric Evaluation

• To determine the accuracy of this metric
• Reputation rating tij is recorded for each
  neighbor j by iterating through each individual
  i in the network
• Later the connection from i to j is removed and
  the reputation rating tij is recorded
• The accuracy is measured as tij-tij

20
TrustMail A Prototype

• Message Scoring System
• Adds reputation ratings to the folder views of a
  message
• Helps sort messages accordingly by the user after
  he sees the reputation ratings
• Highlights the important and relevant messages

21
Conclusion and Future Work

• Our algorithm infers reputation relationships in
  a network
• Benefit - Valid emails from unknown people can
  receive high scores because of the connections
  within the social network
• Future work involves the refinement of the
  algorithm for inferring reputation ratings

22
Conclusion and Future work

• May involve developing and studying the TrustMail
  interface
• The number of ratings received will change with
  the size of a network
• Important issues to be considered
• -Techniques combining best with reputation
  filtering
• - Percentage of messages accurately scored

23
References

• Boykin, P. O. Roychowdhury, V. Personal email
  networks an effective anti-spam tool
  http//www.arxiv.org/abs/cond-mat/0402143,
  (2004).
• http//sites.wiwiss.fu-berlin.de/suhl/bizer/SWTSGu
  ide/
• RDFWeb FOAF The Friend of a Friend
  Vocabulary, http//xmlns.com/foaf/0.1/
• Golbeck, Jennifer, Bijan Parsia, James Hendler,
  Trust Networks on the Semantic Web,
• Richardson, Matthew, Rakesh Agrawal, Pedro
  Domingos. Trust Management for the Semantic
  Web, Proceedings of the Second International
  Semantic Web Conference, Sanibel Island, Florida,
  2003.