Reputation Management for DHT-based Collaborative Environments *



Similar documents
Authenticated Encryption. Jeremy, Paul, Ken, and Mike

Online Load Balancing and Correlated Randomness

ANALYSIS OF ORDER-UP-TO-LEVEL INVENTORY SYSTEMS WITH COMPOUND POISSON DEMAND

Control of Perceived Quality of Service in Multimedia Retrieval Services: Prediction-based mechanism vs. compensation buffers

ERLANG C FORMULA AND ITS USE IN THE CALL CENTERS

How To Write A Recipe Card

Life Analysis for the Main bearing of Aircraft Engines

The example is taken from Sect. 1.2 of Vol. 1 of the CPN book.

An RSA-based (t, n) threshold proxy signature scheme with freewill identities

Modern Portfolio Theory (MPT) Statistics

Protecting E-Commerce Systems From Online Fraud

Personalized Web Search by User Interest Hierarchy

Section 3: Logistic Regression

Modelling Exogenous Variability in Cloud Deployments


The influence of advertising on the purchase of pharmaceutical products

Key Management System Framework for Cloud Storage Singa Suparman, Eng Pin Kwang Temasek Polytechnic

QUANTITATIVE METHODS CLASSES WEEK SEVEN

Mininum Vertex Cover in Generalized Random Graphs with Power Law Degree Distribution

Managing the Outsourcing of Two-Level Service Processes: Literature Review and Integration

Part 2 - Notes on how to complete your application form

Econ 371: Answer Key for Problem Set 1 (Chapter 12-13)

Category 7: Employee Commuting

is knowing the car market inside out.

5 2 index. e e. Prime numbers. Prime factors and factor trees. Powers. worked example 10. base. power

Adverse Selection and Moral Hazard in a Model With 2 States of the World

A Project Management framework for Software Implementation Planning and Management

Bank Incentives, Economic Specialization, and Financial Crises in Emerging Economies

VOL. 25, NÚM. 54, EDICIÓN JUNIO 2007 PP

No 28 Xianning West Road, Xi an No 70 Yuhua East Road, Shijiazhuang.

Facts About Chronc Fatgu Syndrom - sample thereof

Game of Platforms: Strategic Expansion into Rival (Online) Territory

Advantageous Selection versus Adverse Selection in Life Insurance Market

Traffic Flow Analysis (2)

Buffer Management Method for Multiple Projects in the CCPM-MPL Representation

Continuity Cloud Virtual Firewall Guide

User-Perceived Quality of Service in Hybrid Broadcast and Telecommunication Networks

A Secure Web Services for Location Based Services in Wireless Networks*

Term Structure of Interest Rates: The Theories

Rural and Remote Broadband Access: Issues and Solutions in Australia

WORKERS' COMPENSATION ANALYST, 1774 SENIOR WORKERS' COMPENSATION ANALYST, 1769

Architecture of the proposed standard

Tax Collection, Transfers, and Corruption: the Russian Federation at the Crossroads 1)

Non-Linear and Unbalanced Three-Phase Load Static Compensation with Asymmetrical and Non Sinusoidal Supply

Incomplete 2-Port Vector Network Analyzer Calibration Methods

Teaching Computer Networking with the Help of Personal Computer Networks

Constrained Renewable Resource Allocation in Fuzzy Metagraphs via Min- Slack

Panel Discussion: Evolving DoD Security Requirements for Cloud

Scalable Transactions for Web Applications in the Cloud using Customized CloudTPS

ITIL & Service Predictability/Modeling Plexent

Sci.Int.(Lahore),26(1), ,2014 ISSN ; CODEN: SINTE 8 131

A Theoretical Model of Public Response to the Homeland Security Advisory System

C H A P T E R 1 Writing Reports with SAS

FACULTY SALARIES FALL NKU CUPA Data Compared To Published National Data

Lecture 20: Emitter Follower and Differential Amplifiers

Hardware Modules of the RSA Algorithm

Secure User Data in Cloud Computing Using Encryption Algorithms

by John Donald, Lecturer, School of Accounting, Economics and Finance, Deakin University, Australia

Asset set Liability Management for

Entity-Relationship Model

Development of Financial Management Reporting in MPLS

Designing a Secure DNS Architecture

Keywords Cloud Computing, Service level agreement, cloud provider, business level policies, performance objectives.

Intermediate Macroeconomic Theory / Macroeconomic Analysis (ECON 3560/5040) Final Exam (Answers)

Contents Stochastic Ray Tracing

Free ACA SOLUTION (IRS 1094&1095 Reporting)

Contents. Presentation contents: Basic EDI dataflow in Russia. eaccounting for HR and Payroll. eaccounting in a Cloud

A Probabilistic Approach to Latent Cluster Analysis

CPU. Rasterization. Per Vertex Operations & Primitive Assembly. Polynomial Evaluator. Frame Buffer. Per Fragment. Display List.

Logistic Regression for Insured Mortality Experience Studies. Zhiwei Zhu, 1 Zhi Li 2

SPECIAL VOWEL SOUNDS

A DISTRIBUTED REPUTATION MANAGEMENT SCHEME FOR MOBILE AGENT- BASED APPLICATIONS

Sun Synchronous Orbits for the Earth Solar Power Satellite System

(Analytic Formula for the European Normal Black Scholes Formula)

Review and Analysis of Cloud Computing Quality of Experience

An IAC Approach for Detecting Profile Cloning in Online Social Networks

Gold versus stock investment: An econometric analysis

Planning and Managing Copper Cable Maintenance through Cost- Benefit Modeling

A Graph-based Proactive Fault Identification Approach in Computer Networks

CPS 220 Theory of Computation REGULAR LANGUAGES. Regular expressions

Parallel and Distributed Programming. Performance Metrics

ESA Support to ESTB Users

The international Internet site of the geoviticulture MCC system Le site Internet international du système CCM géoviticole

union scholars program APPLICATION DEADLINE: FEBRUARY 28 YOU CAN CHANGE THE WORLD... AND EARN MONEY FOR COLLEGE AT THE SAME TIME!

Cookie Policy- May 5, 2014

benefit is 2, paid if the policyholder dies within the year, and probability of death within the year is ).

Abstract. Introduction. Statistical Approach for Analyzing Cell Phone Handoff Behavior. Volume 3, Issue 1, 2009

Traffic Information Estimation Methods Based on Cellular Network Data

IT09 - Identity Management Policy

Mathematics. Mathematics 3. hsn.uk.net. Higher HSN23000

Use a high-level conceptual data model (ER Model). Identify objects of interest (entities) and relationships between these objects

Transcription:

Rputaton Managmnt for DHT-basd Collaboratv Envronmnts * Natalya Fdotova, Luca Vltr Unvrsty of Parma, Italy Abstract Ths artcl addrsss a problm of ntgraton of rputaton managmnt mchansms and lookup procsss n DHT-basd pr-to-pr (PP) ntworks wth collaboratv natur. W prsnt a possbl scnaro of applcaton of som rputaton tchnqus to lookup and data rtrval procsss n a PP ntwork basd on th popular Kadmla algorthm. Unlk most of xstng rputaton managmnt systms for DHT-basd PP nvronmnts, w don t us a numbr of succssful downloads as th man nstrumnt to valuat a trustworthnss of prs, but w propos to consdr any typ of ntractons btwn nods. Ths choc s motvatd by th fact that fl sharng s not a unqu srvc supportd by DHT-basd PP systms. Thr ar othr applcaton aras for PP (.g. collaboratv and dstrbutd computng) whr t s mportant to consdr such aspcts as a rsk factor, a stay on-ln tm or a numbr of rqusts wthout rply. Kywords-componnt Pr-to-pr, Dstrbutd Hash Tabls, rputaton mchansms, scurty. I. INTRODUCTION: BACKGROUND AND MOTIVATION Dstrbutd Hash Tabls (DHT) ar buldng block for ralzaton of many pr-to-pr (PP) applcatons and srvcs, such as dstrbutd fl systms, fl sharng, nstant mssagng, collaboratv and dstrbutd computng. DHTs provd a vry ffcnt nstrumnt for lookup, routng, data storag and sharng mchansms [1,]. Howvr, DHT-basd PP ntworks rprsnt a partcular nvronmnt vulnrabl to som spcfc thrats and attacks. Gnrally, ths attacks ar causd by malvolnt bhavour of som ntwork nods and thy ar amd at routng and lookup procsss [3]. Th slf-organzng natur of DHTs nabls som countrmasurs aganst svral ffcts of such malcous actvty. For nstanc, n th cas of tratv lookup (Kadmla, Chord) ncorrct lookup routng attacks can b avodd by chckng th progrss of lookup at ach stp; storag and rtrval attacks can b prvntd by rplcaton of fls usng multpl hash functons; Sybl attacks cannot b xcludd n such nvronmnt, but a lookup ffcncy can b mprovd by paralll routng (ssung svral lookup rqusts at a tm) [4]. * Ths work has bn partally supportd by th Italan Mnstry for Unvrsty and Rsarch (MIUR) wthn th projct PROFILES undr th PRIN 006 rsarch program. Anyway, ths countrmasurs ar momntary solutons that don t rsolv th problm of pollutng routng tabls by malcous contacts. Thrfor, t s rasonabl to us gand xprncs to rndr a ntwork communty mor scur. Onc a malvolnt actvty has bn dtctd, rputaton tchnqus should b appld n ordr to clar routng tabls from contacts that hav vncd malcous or nconsstnt bhavour to avod thm n th futur. Rcntly, a numbr of rputaton managmnt tchnqus for PP ntworks has bn proposd by dffrnt rsarchrs. In our prvous work [5] w prsntd a dtald analyss of applcablty of svral xstnt rputaton valuaton tchnqus as protcton from som typs of attacks n DHTbasd PP ntworks. Th analyzd tchnqus ar not dsgnd for DHT-basd PP ntworks, so non of thm rprsnts a unvrsal soluton for such systms. At th sam tm, dffrnt rputaton managmnt nstrumnts usd by ths tchnqus could b a qut ffctv n som partcular cass n a DHT nvronmnt. Normally, rputaton tchnqus proposd spcfcally for DHT-basd ntworks ar havly basd only on valuaton mchansms of succssful and unsuccssful downloads, snc thy rgard only fl sharng PP applcatons [6,7]. Howvr, as w hav mntond abov, PP tchnology also supports nstant mssagng, collaboratv applcatons, dstrbutd computng, tc. In th last fw yars PP systms hav bn succssfully usd for sharng computaton undr varous dstrbutd computng projcts lk FghtAIDS@Hom, Gnom@Hom, St@Hom, Untd Dvcs Cancr Rsarch Projct and othrs [8]. Ths projcts rprsnt a publc-rsourc computng that rls on prsonal computrs wth xcss capacty, such as dl CPU tm, to rsolv som complx rsarch problms. Publc-rsourc computng s an aspct of th pr-to-pr paradgm, vn f t uss a grd tchnology to ralz ts tasks. Currntly, such systms approach a DHT natur. Som stps of computatonal procsss bcom compltly ndpndnt from cntral srvrs: calculatons rsults of som nod ar stord n th ntwork and rtrvd by a succssor that us thm n ts own part of th task; f a pr lavs a ntwork whl procssng a work unt, th work unt s vntually snt to anothr pr that bcoms rsponsbl for t (lk n DHT data storag systms whn a nod bcoms rsponsbl for rsourcs of som fald nod f thr dntfrs ar consdrd as th closst to ach othr). 978-1-444-1785-8 316

Anothr typ of systms that us DHT prncpls ar collaboratv applcatons for data storag and dtng by svral gographcally dstant work groups. Such systms should provd a rapd and scur data xchang btwn dffrnt systm unts and possblty of tam-work n raltm and transparnt mod. Th systms dscrbd abov rprsnt actv dstrbutd collaboratv nvronmnts, whr vry ntracton btwn prs s mportant and, as L. Lamport sad, th falur of a computr you ddn't vn know xstd can rndr your own computr unusabl [0]. So, n ths systms a numbr of succssful downloads cannot b a suffcnt nstrumnt for rputaton managmnt. In ths cas paramtrs charactrzng a communty contxt (rsk factor, numbr of lookup rqusts wthout rspons, numbr of jon and lavs for a nod, offln status tm) ar also rqurd. In ths work w consdr th possblts of ncorporaton of rputaton mchansms n DHT-basd routng and lookup procsss. W propos a soluton that combns dffrnt nstrumnts offrd by som rputaton managmnt tchnqus analyzd n our prvous work, applyng thm to lookup and data rtrval procsss n a PP ntwork basd on th popular Kadmla algorthm. II. PREVIOUS WORK Takng n consdraton som partcularts of dployng rputaton mchansms n a DHT sttng, w ntroducd th followng applcablty crtra for rputaton mchansms: 1. tchncal ralzablty n ovrlay ntworks;. avalablty of ndvdual rputaton valuaton nstrumnts. Th frst crtron s crtcal for rputaton valuaton tchnqu wth crdntal and polcy lmnts [9]. In crdntal and polcy basd trust managmnt systms prs us a st of crdntals and polcs to dtrmn whthr som unknown pr can b trustd or not. Obvously, n ths cas th prsnc of som sort of crtfcaton authorty s rqurd. Such tchnqus oftn rqur a cntral srvr for storng and dstrbutng rputaton nformaton. Thrfor, crdntal and polcy basd mchansms ar to b appld n cntralzd systms wth a hrarchcal structur. In th cas of compltly dstrbutd DHT ntworks w nd slf-polcng tchnqus provdng mchansms that can b ralzd n dcntralzd nvronmnt usng mans proposd by an ovrlay nfrastructur, lk pur rputaton mchansms. Th scond crtron s rprsntd by som dffrnt paramtrs, such as: possblty to provd rcommndatons; possblty to wgh rcommndatons,.. rcommndatons from dffrnt prs hav dffrnt lvls of trustworthnss; rsponsblty for th bhavor of rcommndd ntts; valuaton of a communty contxt (an avrag lvl of vulnrablty of th ntwork nvronmnt and a lvl of coopraton btwn prs); ncntvs for fdback complaton. So, a sngl nod should b provdd wth all ncssary nstrumnts to analyz and ndpndntly valuat rputaton and rlablty of othr prs. Tabl 1 summarzs rsults of our comparatv analyss takng nto account data rgardng all th basc charactrstcs of th analyzd rputaton modls. Concrnng th frst applcablty crtron, most of th analyzd tchnqus can b subsumd undr th rputaton basd catgory. NICE [15] and DCRC/CORC [19] ralz som crdntal and polcy lmnts: dgtal sgnaturs of cooks n NICE, pr dntfcaton by rputaton computaton agnts (RCAs) usng publc ky n DCPC/CORC. Poblano [14] and XREP [16] also nvolv som mchansms wth a cntralzd natur. Ths fact rprsnts som dffcults for applcaton of ths tchnqus to compltly dstrbutd DHT-basd ntworks. As to th scond crtron, all ths tchnqus hav dffrnt compltnss dgrs. PrTrust [1] and Fuzzy Modl [11] rprsnt th most complt tchnqus, as thy ralz almost all possbl mchansms for valuaton of a pr s trustworthnss. W can conclud that non of ths tchnqus n ts pur form rprsnts a sutabl soluton for DHT-basd PP ntworks. Th soluton w prsnt s a combnaton of dffrnt nstrumnts offrd by th analyzd modls. W propos to apply ach of ths nstrumnts whn t s consdrd as th most ffcnt on for a crtan stuaton. III. INTEGRATION OF REPUTATION MECHANISMS AND LOOKUP PROCESSES A. Lookup and rputaton mchansms In DHT-basd systms (CAN, Chord, Pastry, Kadmla) a group of dstrbutd hosts collctvly manags a mappng from kys to data valus, wthout any fxd hrarchy, and wth a vry lttl human assstanc [1]. Th bas of such systms s a routng tabl-basd lookup srvc. If a lookup ntator dosn t fnd n ts routng tabl an ID of a nod rsponsbl for som dsrd ky, t snds a lookup qury to a nod t consdrs th closst to ths ky among all ts contacts. Th last on should rply wth an IP addrss of th nxt hop (tratv lookup) or forward th qury to th nxt closst nod (rcursv lookup). Applyng opportun 317

TABLE I. APPLICABILITY ANALYSIS SUMMARY TABLE Paramtr Modl Supportng Trust n Vrtual Communts Pr Trust Prsonalzd Trust Modl Fuzzy Modl for contxt-dpndnt Rputaton Poblano Rputaton Valu Scal 4 possbl lvls Normalzd Normalzd Normalzd Possblty to Provd Rcommndatons Possblty to Wgh Rcommndatons Rsponsblty for th Bhavor of Rcommndd Entts Transacton consdrato n (typ, quantty, data, dmnsons) Evaluaton of a communty contxt Incntv to fdback complatons Ys Ys No No No No No Ys¹ Ys² No Ys Ys Ys No Ys No No No No No No Ys Ys Ys Ys³ Ys No No 6 possbl No No No No No No No lvls [-1; 4] NICE Normalzd Ys 4 No No No No No No XREP Bnary 5 Ys 6 No No No No No Ys Crdntal and polcy lmnts Sporas and Hstos Bta DCRC/CORC From 0 to 3000 xclusv Normalzd Nonngatv Ys¹ Ys² No No No No No Ys¹ Ys² No No No No No No No No Ys No No Ys 1 fdbacks compld by othr usrs ar consdrd 3 only th vnts numbr s consdrd 5 - a bnary valu scal s not oblgatory, t should just b ndcatd an ntrval of valus usd rgards an ntty complng a fdback 4 rcommndatons rgardng som dtrmnd pr 6 n th form of a vot mchansms for vrfyng lookup progrss, th quryng nod can also mak a concluson about honsty of th nods partcpatng n th lookup procss, assgnng to thm th corrspondng rputaton valus. Analogcally, a nod that honstly shars ts rsourcs wth othr nods gts rputaton ponts, and a nod dnyng th xstnc of data t s rsponsbl for (storag and rtrval attacks), loss thm. So, ach nod of th ntwork aftr vry contact wth anothr nod assgns a nw rputaton valu to th contactd pr dpndng on th ntracton rsults. All th assgnd rputaton valus ar to b stord by th quryng nod and should b consultd bfor contactng corrspondng nods agan. Morovr, ths rputaton valus ar usd as rcommndatons that ach nod xchangs wth othrs, whn rspondng to an tratv lookup qury. It s mportant to not that n th cas of rcursv lookup t s problmatc to apply vrfyng mchansms at ach hop. So, to avod forwardng qurs to malcous prs, nods should also chck rputaton valus of potntal dlgats whn choosng nxt stp of lookup routng nstad of smply consdrng thr closnss to a dsrd rsourc. If som nod s dclard untrustworthy on th bas of gand rputaton ponts, t should b gnord by good prs durng th futur lookup and routng procsss,.. t should b smply consdrd and tratd as a nod that has lft th ntwork (a nod wth off-ln status). Ths mchansm dosn t affct th stablty of a DHT-basd ovrlay du to th multpl data rplcaton and kys rassgnmnt schm usd n such nvronmnt. All data rplcas ar unformly dstrbutd on th ntwork btwn svral prdfnd rsponsbl nods. So, t s always possbl to fnd anothr nod nabld to provd th sam rsourc. B. Proposd scnaro In ths scton w prsnt a possbl scnaro of applcaton of som rputaton mchansms provdd by th analyzd tchnqus. As an nvronmnt for our scnaro w chos a ntwork basd on Kadmla DHT protocol, wdly usd by a numbr of PP platforms (Donky, BtTorrnt, tc). Th proposd combnaton of th rputaton nstrumnts ncluds: rsk valuaton mthod provdd by PET modl [13]; 318

rsourcs and srvnt rpostors from XREP modl [16]; dbt-crdt basd rputaton computaton modl (DCRC) [19]. Ths scnaro rprsnts a stuaton n whch a pr jons a ntwork th frst tm and ntats a lookup procss for a data fl wth a crtan ID. Snc t s a nw nod for ths ntwork, t has no da about trustworthnss of othr nods. As rputaton s an accumulatv valu, t s not possbl to valuat somon s rputaton just aftr th frst contact. Howvr, t s possbl to dfn a lvl of vulnrablty of th ntwork nvronmnt on th bas of rsults of th frst xprncs usng th approprat nstrumnt offrd by PET. PET modl drvs th trustworthnss valu T from two componnts: rputaton factor R and rsk factor R wth dffrnt wghts (ncdnc) W and R W R rspctvly. Th trustworthnss valu s dfnd as follows: T T W R + W R ( α,1 α ) ( R, (1 R )),0 α 1 (1), R R whr W R α and W R 1 α, and th valus of T, R and R ar all. In our scnaro, aftr th frst contact T R, as th nod dosn t hav suffcnt data to valuat th rputaton factor, but t s ncssary for t to dfn a lvl of vulnrablty of th ntwork t has jond. Th rsk valu n PET modl s calculatd by th formula: ( N h( )) B, N, L R h( B) N j j G, B, N. L (), whr G, B, N, L ar four lvls of qualty (Q) of srvcs provdd by a pr (t s applcabl to any typ of ntracton btwn prs: laboraton and forwardng of qurs durng a lookup procss, provdng rsourcs, tc.): G Good, L Low Grad, N No Rspons, B Byzantn Bhavour; N s th numbr of srvcs (ntractons) provdd wth qualty ; h s a map functon from Q to a scor for on ntracton btwn nods,.. t shows how many rputaton ponts a nod has gand or lost at th nd of on ntracton: S1, Q G, S1 > 0 S, Q L, S < 0, and S h( Q) S 3, Q N, S3 < S S 4, Q B, S 4 < S3 > S 1 (3). Fg.1. Look-up procss basd on Kadmla algorthm n trms of PET modl So, w can s that msbhavng rducs a rputaton valu fastr than a good bhavour ncrass t. No Rspons s consdrd as a bad bhavour hr. It hlps to avod too frqunt and long lavng n PP ntworks: t s not convnnt for a pr to b off-ln for a long prod, bcaus all rqusts snt to th nod durng ts absnc wll rst wthout rspons, sgnfcantly rducng ts rputaton. It s xpland by a hgh mportanc of th coopraton componnt for collaboratv nvronmnts: a nod that dosn t partcpat n routng and lookup mchansms by procssng and forwardng rqusts from othr nods cannot b postvly valuatd by th communty. Th rsk valu R s normalzd to th worst cas, that rprsnts a stuaton whn all srvcs rcvd by a pr n a crtan tm ntrval ar Byzantn srvcs. In our xampl w assgn to S1, S, S3 and S4 th valus 1, -, -3 and -4 rspctvly. In Kadmla, dntfrs of nods consst of 160 bt. Hr w prsnt a smplfd modl of Kadmla s bnary tr wth a lttl numbr of sub-trs and lavs. As sn n fgur 1, th nod wth prfx 0011 ntats th procss of look-up for som rsourc, sndng FIND_VALUE RPC to two nods rsdng n two dffrnt sub-trs wth prfxs 0111 and 1011. Th frst nod dosn t rspond (th on-drctonal dashd arrow). Th scond nod rturns a trpl <IP addrss, UDP port, Nod ID> for th nod 1101. Applyng th look-up progrss vrfcaton mchansm, th rqustor concluds that th Nod ID snt by th scond nod s rally closr to th ky. Accordng to PET modl our rqustor assgns to ths contact th qualty valu G, and to th frst contact th valu N. Thn, t snds FIND_VALUE RPC to nod 1101, that rturns a trpl contanng Nod ID 1110, but wth a lttl dlay. In ts turn, 1110 rpls wth Nod ID 11110, that stors th dsrd data. 11110 rturns th stord valu. Havng controlld at ach stp th look-up progrss, th rqustor assgns to nods 1101, 1110 and 11110 th qualty valus L, G and G rspctvly. 319

Th rsk valu n our cas s: NL S + NN S R S ( N + N + N 4 1 ( ) + 1 ( 3) ) 4 (3+ 1+ 1) 3 G L N 0,5 Informaton, rgardng th qualty of rsourcs and th rlablty of th prs obtand durng ths look-up procss, could b stord n th rsourc and srvnt rpostors accordng to th XREP modl [16]. Th rpostors rprsnt two tabls wth th followng data structurs: <rsourc_ d, valu>; <srvnt_d, num_plus, num_mnus>. Th frst tabl assocats wth ach rsourc dntfr (rsourc_d) a valu that dfns th qualty of th corrspondng rsourc. XREP s authors don t prcs a typ of data rprsntng qualty valus allowng th lbrty of ntrprtaton: t may b a numrc valu, or t may b smply dfnd as good or bad. Th data of a srvnt rpostory contan IDs (srvnt_d) of contactd prs and corrspondng numbrs of succssful (num_plus) and unsuccssful (num_mnus) transactons ffctuatd wth ths prs. As n our cas w us DCRC modl, t maks sns to substtut th num_plus and num_mnus wth data rgardng th total numbr of uploadd and downloadd mgabyts of contnt for a crtan pr,.g. mb_up and mb_down. To calculat th Qury-Rspons Crdt (QRC) paramtr, for ach contactd nod th total numbr of qurs addrssd to a nod and a numbr of qurs procssd and forwardd by a nod should b stord. Adaptng th data structur of th srvnt rpostory to our cas, th lattr tabl bcoms: <srvnt_d, mb_up, mb_down, num_qury, num_rply>. Accordng to th prvous xampl, th nod 0011 stors th followng data n ts srvnt rpostory: srvnt_d mb_up mb_down num_qury num_rply 0111 0 0 1 0 1011 0 0 1 1 1101 0 0 1 1 1110 0 0 1 1 11110 0 0 1 1 In trms of tchncal ralzaton ths mchansm rprsnts a qut smpl soluton: a smpl countr s ssntally rqurd. Accordng to DCRC modl [19], a rputaton valu of a pr s dfnd by crdts t gans or loss durng a crtan prod ntractng wth othr prs. Th total numbr of rputaton ponts can b calculatd by th followng formula: nu UCl nd DDl + t f R n QR + SC (4), q l m whr: QR s th numbr of ponts gand by a nod for ach procssd qury; n s th total numbr of qurs procssd by a pr; q n u s th numbr of uploads facltatd by a pr; n d s th numbr of downloads prformd by a pr; t s th prdfnd tm factor (n hours) that dtrmns a f tm ntrval durng whch th dscrbd ntractons hav bn prformd; UC and l DD ar th upload crdt and download dbt for m fls l and m rspctvly. In ts turn th UC s dfnd as follows: s bw UC (5), f b whr: s s th sz of an uploadd fl (n mgabyts); bw s th bandwdth avalabl (n mgabyts); f s th fl sz factor that dtrmns how many mgabyts of data transfr ncrass th rputaton scor by a unt; b s th bandwdth factor that classfs prs on th bas of bandwdths thy hav at dsposal. Th DD for a download of a rsourc of sz s s gvn by: s bw DD (6), f b whr bw s th bandwdth of a pr from whch a download s prformd. Th Sharng Crdt (SC) for a pr that shars n fls durng som prdfnd tm ntrval s gvn by: n s j SC (7), j f whr, s j s th sz of jth fl. Lt s calculat th rputaton valu of th nod 11110 usng th data stord by 0011 at th rpostory aftr th consdrd look-up procss. Lt th sz s of th fl downloadd by th nod 0011 from 11110 b 0 MB, th fl sz factor f MB, th avalabl bandwdth bw 6MB, and th bandwdth factor b MB. Thn, th UC of 11110 aftr ths ntracton s: 0 6 UC 30 For smplcty, lt th numbr of ponts gand by a pr for ach qury procssd QR1. Lt s suppos that th total sz of th rsourcs shard by 11110 nod s 500 MB and th prdfnd tm ntrval s 1 hour. Hnc, SC50. 30

Thn, at n q 1 and DD0, th total numbr of rputaton ponts gand by th pr 11110 aftr th ntracton n quston s: R QRC + UC DD + SC 81. tot tot tot tot IV. CONCLUSION Th rsults of th applcablty analyss from our prvous work show that non of th consdrd rputaton tchnqus n ts pur form rprsnts a unvrsal and sutabl soluton for DHT-basd PP nvronmnt. So, w hav proposd a soluton that conssts n applcaton of a combnaton of dffrnt rputaton mchansms provdd by som analyzd tchnqus. It s motvatd by th fact, that xstng rputaton valuaton tchnqus dsgnd spcfcally for DHT-basd ntworks manly for fl-sharng PP applcatons ar basd only on a numbr of succssful and unsuccssful downloads. Howvr, although ths mchansm can b consdrd as an ndcatv componnt of rputaton valuaton, t cannot b a suffcnt nstrumnt for rputaton managmnt n collaboratv nvronmnts (.g. dstrbutd computng and data dtng). In collaboratv nvronmnts t s also vry mportant to consdr possbl rsks and varous paramtrs rgardng a communty contxt, bcaus th cost of a mstak, causd by malcous actvty n such ntworks s ncomparably hghr than n fl-sharng systms. Just a fw unrlabl prs that hav not bn dscardd from routng tabls n tm can ntrrupt a long chan of calculatons. In ths work w hav prsntd a possbl scnaro of applcaton of th proposd soluton to a PP ntwork basd on Kadmla DHTs, xtractng som rputaton valuaton nstrumnts from th modls analyzd bfor and adaptng thm to partcularts of DHT-basd nvronmnt. Th proposd soluton rprsnts an ndvdual mchansm of rputaton managmnt for a sngl pr basd on ts own xprnc. W can conclud that, th rsk calculaton mthod usd n PET modl hlps to dfn a lvl of vulnrablty of an unfamlar nvronmnt, whl DCRC tchnqu rprsnts a an objctv mthod of rputaton valuaton basd on ponts gand by a pr du to ts collaboraton wth th communty. In ts turn, th rpostory mchansm s a smpl and ffcnt soluton for systmatzaton of th data ncssary for rputaton ponts calculaton. All of ths nstrumnts can by asly adaptd to DHTbasd nvronmnt. Ths xampl dmonstrats that th mchansms usd n our scnaro succssfully complmnt ach othr, vn f thy wr xtractd from thr dffrnt rputaton valuaton modls. W ar currntly workng toward a softwar mplmntaton of th dscrbd scnaro and othr rputaton mchansms n Kadmla ntwork. REFERENCES [1] D. Doval, D. O Mahony, Ovrlay Ntworks: A Scalabl Altrnatv for PP, IEEE Journal on Intrnt Computng, Vol.7, No.4, pp. 79-8, August 003 [] H. Balakrshnan, M. F. Kaashok t al., Lookng Up Data n PP Systms, Communcatons of th ACM, Vol. 46, No., pp.43-48, Fb. 003. [3] E. St, R. Morrs, Scurty consdratons for Pr-to-Pr Dstrbutd Hash Tabls, n Procdngs of th Frst Intrnatonal Workshop on Pr-to-Pr Systms (IPTPS 0), Cambrdg, Massachustts, March 00 [4] D. Stutzbach, R. Rja, Improvng Lookup Prformanc ovr a Wdly-Dployd DHT, n Procdngs of 5th IEEE Intrnatonal Confrnc on Computr Communcatons, INFOCOM 06, Aprl 006 [5] N. Fdotova, M. Brtucc, L. Vltr, Rputaton Managmnt Tchnqus n DHT-basd Pr-to-Pr Ntworks, n Procdngs of ICIW 07, May 007 [6] S. D. Kamvar, M.T. Schlossr, H. Garsa-Molna, Th Egn Trust Algorthm for Rputaton Managmnt n PP Ntworks, n Procdngs of th 1 th Intrnatonal World Wd Wb Confrnc, May 003 [7] S. Y. L, O-H. Kwon, J. Km, S. J. Hong, A Rputaton Managmnt Systm n Structurd Pr-to-Pr Ntworks, n Procdngs of 14 th IEEE Intrnatonal Workshops on Enablng Tchnologs: Infrastructurs for Collaboratv Entrprs (WETICE 05) [8] V. Martns, E. Pactt, P. Valdurz, Survy of data rplcaton n PP systms, rsarch rport, INRIA, Dcmbr 006 [9] G. Suryanarayana, R.N. Taylor, TREF: A Thrat-cntrc Comparson Framwork for Dcntralzd Rputaton Modls, ISR Tchncal Rport UCI-ISR-06-, January 006 [10] A. Abdul-Rahman, S. Hals, Supportng Trust n Vrtual Communts, In Procdngs of th 33 rd Hawa Intrnatonal Confrnc on Systm Scncs, Mau, Hawa, January 000. [11] V. Grshchnko, A fuzzy modl for contxt-dpndnt rputaton, Trust, Scurty and Rputaton Workshop at ISWC 004, Hroshma, Japan. [1] L. Xong, L. Lu, A Rputaton-Basd Trust Modl for Pr-to-Pr Ecommrc Communts, Proc. IEEE Conf. E-Commrc (CEC '03), Jun 003. [13] Z. Lang, W. Sh, PET: A PErsonalzd Trust Modl wth Rputaton and Rsk Evaluaton for PP Rsourc Sharng, In Procdngs of th 38 th Hawa Intrnatonal Confrnc on Systm Scncs, 005. [14] R.Chn, W. Yagr, Poblano: A Dstrbutd Trust Modl for Pr-to- Pr Ntworks. Tchncal rport, Sun Mcrosystms. http://www.jxta.org/docs/trust.pdf. [15] L, S., Shrwood, R., t al., Coopratv pr groups n NICE, IEEE Infocom, San Francsco, USA, 003. [16] Daman. E., d Vmrcat S., t al., A Rputaton-Basd Approach for Choosng Rlabl Rsourcs n Pr-to-Pr Ntworks, 9th ACM Confrnc on Computr and Communcatons Scurty, Washngton DC, 00. [17] Zachara, G. and Mas, P., Collaboratv Rputaton Mchansms n Elctronc Marktplacs, 3nd Hawa Intrnatonal Confrnc on Systm Scncs, Hawa, 1999. [18] A. Josang, R. Ismal, Th Bta Rputaton Systm, 15th Bld Elctronc Commrc Confrnc, Bld, Slovna, 00. [19] M. Gupta t al., A Rputaton Systm for Pr-to-Pr Ntworks, Thrtnth ACM Intrnatonal Workshop on Ntwork and OpratngSystms Support for Dgtal audo and Vdo. Montry, Calforna, 003. [0] http://rsarch.mcrosoft.com/usrs/lamport/pubs/pubs.html 31