Your Federal Quarterly Tax Payments are due April 15th Get Help Now >>

Reputation Management Ppt by kue20725

VIEWS: 638 PAGES: 1

Reputation Management Ppt document sample

More Info
									Reputation Management for Web Services
Athman Bouguettaya, and Mohamed Eltoweissy
   p          g                p
http://www.eceg.cs.vt.edu/WS-reputation/index.htm              NSF Grant CNS-0627469


 Reputation Model
    W use reputation as a means t establish t t b t
    We            t ti                                            i i  k       Web
                                   to t bli h trust between a priori unknown W b
                               p                                           p
  services. The model is cooperative in that Web services share their experiences
                      ratings,
  through feedback ratings which are then used for reputation assessment in absence
   f        t li d trusted
  of any centralized t t d agency.
    The service providers are rated
       g                     y
  along different Quality of Service
                       (e.g., availability,
  (QoS) attributes (e g availability
  reliability, accessibility, etc).
  reliability accessibility etc) Based
  on several metrics that we have
  defined the service consumers
  defined,
  independently        aggregate        the
                   g
  available ratings. This ensures that
  each consumer has a “local          local
               view         providers.
  reputation view” of the providers

  An    end to end
        end-to-end   solution   that
  considers all facets of reputation
       g         p p
  management is proposed.
                                   Approach and Impact
                                   A      h dI       t
          New approach
          N          h                                            Research Impact
                                                                  R      hI     t
    Reputation
  • Rep tation Bootstrapping                     Fairness for new service’ selection
                                                •F i      f           i ’ l ti
  • Rater Credibilities                         • Attenuation of effects of dishonest ratings
  • Reputation Metrics                          • Accurate reputation assessment
  • Reputation-based Selection                  • Optimization of service selection

 Technical Description
        p p                          g
    The proposed model is ratings-based. However, in situations where ratings are not g
 available, we incorporate probability (using hidden markov models) to predict service
 reputation.        ratings, rater s
 reputation Past ratings rater’s propensity to be dishonest and current ratings are used to
    ti t        t ’       dibilit thus     dif i the “majority is th it ” l that is
 estimate a rater’s credibility, th modifying th “ j it i authority” rule, th t i mostly        tl
                   g             y
 followed in ratings-based systems.
                                                   ontology based               model,
    The service space is organized using an ontology-based community model in which
 consumers and providers are required to register with a community of interest In    interest.
      i i        t ti to               i            id        l ti that includes
 assigning reputation t new services, we provide a solution th t i l d as a parameter          t
                                      y                     y
 the current rate of dishonesty in the community. For services that are part of          p
 compositions we use two types of reputations: (i) individual-based (ii) composition-based
 compositions,                                           individual based    composition based.
 Results
   Highly                     i                  f     id      h is l fair       i i
 • Hi hl accurate reputation assessment of providers that i also f i to existing and new   d
 services
 • Better results than similar state-of-the-art solutions
 • Collusion/Bad mouthing are avoided
          p                         p         y g
 • In composition, the actual culprit is likely to get the blame
                             NSF Cyber Trust Principal Investigators Meeting
                                          March 16-18, 2008
                                            New Haven, CT                             Replace this logo with
                                                                                       your school’s logo

								
To top