A notoriety based trust administration in shared system frameworks

2657 days ago, 1100 views
PowerPoint PPT Presentation
Plot. Associate to-peer(P2P) systems overviewRelated workProposed approachExperiments. Shared (P2P)

Presentation Transcript

Slide 1

A notoriety based trust administration in distributed system frameworks Natalia Stakhanova, Sergio Ferrero, Johnny Wong, Ying Cai Department of Computer Science Iowa State University Ames, Iowa, USA

Slide 2

Outline Peer-to-peer(P2P) systems diagram Related work Proposed approach Experiments

Slide 3

Peer-to-associate (P2P) systems review P2P arrange - an overlay system of companions trading assets basic uses: record sharing, conveyed figuring, texting sorts: Centralized (Napster) Central ordering server contains data about all associate's shared documents Decentralized (Gnutella, Freenet) No focal ordering, all associates are equivalent Very well known Hybrid (KaZaA, FastTrack) Supernodes keep up file of documents shared by their neighborhood peers

Slide 4

P2P systems outline Differences with customary systems Highly dynamic self-ruling companions peers leave & join the system whenever shared capacity Peers go about as servers and customers

Slide 5

P2P security dangers Denial-Of-Service assaults (DoS) Decentralized P2P systems (Gnutella) Virus circulation Dishonest transfer Unauthorized access to data Goal: correspondence with trusted associates just

Slide 6

Reputation-based approach Natural instrument for selecting trusted accomplices for correspondence restrict correspondence with inconsistent companions Most normally utilized

Slide 7

Related work Centralized methodologies Debit-Credit Reputation Computation (DCRC) composition Each companion tracks its own positive commitment utilizing credit-charge component Reputation Computation Agent (RCA) occasionally gathers notorieties Decentralized methodologies NICE Reputation is in type of treats which express associate's fulfillment about the exchanges If no treat is discovered data is asked for from P2PRep Reputation of the associate depends on other associate's conclusion Request associate's sentiment on one's notoriety through surveying convention Others Daswani and Garcia-Molina " s pattern for dispensing assets reasonably Traffic administration in view of l oad-adjusting strategies DoS assaults just

Slide 8

Factors to be considered in notoriety based approach Extensive movement in Gnutella-like P2P system Storage focal nearby Cooperation of different associates System overhead

Slide 9

Proposed approach Reputation estimation is based the observed action of the associated peers evaluating the notoriety of the associates before tolerating activity from different associates if movement is acknowledged redesign notoriety of companions included Decentralized - notorieties are put away and oversaw locally

Slide 10

Contribution of our approach Fully decentralized model Requires no collaboration for notoriety calculation On request counts Lightweight – little framework overhead

Slide 11

Reputation count Peer's notoriety demonstrates its commitment to the working of the P2P organize Four variables deciding notoriety: Resource seek Resource transfer Resource download Traffic breadth Factors = activities Bad activities Good activities

Slide 12

Resource look eagerness of an associate to forward movement employ " trailer " as an option to Query message every companion that advances the question adds its ID to the " trailer " when companion shapes QueryHit, it exchanges a " trailer " from Query to QueryHit peer started an inquiry gets QueryHit with " trailer " and overhauls notorieties

Slide 13

Resource transfer Indicates another associate " s enthusiasm for the common asset Completely transferred record is a fruitful transfer or great activity

Slide 14

Resource download mirrors the nature of the downloaded data User chooses if download was effective

Slide 15

Traffic breadth assess the activity stack originating from every single associated companion in light of the normal load is broad on the off chance that it surpasses the normal sum by a client pre-characterized edge L cK - current load from companion k t - edge n - number of associated associates l j - number of bytes sent by associate j n L cK > ∑ l j/n * t j=1

Slide 16

Reputation count Reputation esteem (trust score) is a percent of awful activities happened amid a timeframe R i = BA i/TA i R i - trust score of companion i TA i - add up to number of considered activities for this companion i BA i - number of terrible activities for this companion i

Slide 17

Trust edges show peer " s trust strategy percent of awful activities satisfactory by the companion

Slide 18

The correspondence between trust limits and trust score Example: trust score falls in scope of " normal " -> x 1 – (R i – x 2 ) Computations: 30-(13-4) = 21 21% of associate " s activity is acknowledged inside period k. Given : R i =13 x 1 =30 x 2 =4

Slide 19

P2P customer … Security Manager Reputation Manager Reputation storehouse Internet Connection Engine Experiments: framework outline usage depended on Phex variant, a java-based Gnutella customer

Slide 20

Experimental setup Network : 3 P2P customers set up as Ultrapeers companion limit - 20 inquiries for every day and age k k=5 sec Extensive movement edge t=1.7 Trust edges x 1 =20 x 2 =5 Initial notoriety values for associates were set up physically

Slide 21

Scenario 1 Decrease of full notoriety when peer P1 begins " acting " malignantly

Slide 22

Scenario 2 Reputation pick up when companion begins " acting " legitimately

Slide 23

Conclusion We have proposed notoriety based trust administration display for P2P systems approach is decentralized requires no associates " participation utilizes just on-request figurings

Slide 24

Future work Enhancement of the model through client profiling procedures abnormality location