Trust and reputation models among agents

87
Trust and reputation models among agents Nick Bassiliades Intelligent Systems group, Software Engineering, Web and Intelligent Systems Lab, Dept. Informatics, Aristotle University of Thessaloniki, Greece Invited talk at 6th International Conference on Integrated Information Sep 19-22, 2016, Athens, Greece

Transcript of Trust and reputation models among agents

Page 1: Trust and reputation models among agents

Trust and reputation models among agents

Nick Bassiliades

Intelligent Systems group, Software Engineering, Web and Intelligent Systems Lab,

Dept. Informatics, Aristotle University of Thessaloniki, GreeceInvited talk at

6th International Conference on Integrated InformationSep 19-22, 2016, Athens, Greece

Page 2: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 2

TALK OVERVIEW Introduction Review of State-of-the-Art on Trust / Reputation Models Centralized approaches Distributed approaches Hybrid approaches

Presenter’s work on Trust / Reputation Models T-REX HARM DISARM

Summary / Conclusions

IC-ININFO, Sep 19-22, 2016

Page 3: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 3

TRUST AND REPUTATION Agents are supposed to act in open and risky

environments (e.g. Web) with limited or no human intervention

Making the appropriate decision about who to trust in order to interact with is necessary but challenging

Trust and reputation are key elements in the design and implementation of multi-agent systems

Trust is expectation or belief that a party will act benignly and cooperatively with the trusting party

Reputation is the opinion of the public towards an agent, based on past experiences of interacting with the agent

Reputation is used to quantify trustIC-ININFO, Sep 19-22, 2016

Page 4: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 4

TRUST / REPUTATION MODELS Interaction trust: agent’s own direct experience from past

interactions (aka reliability)- Requires a long time to reach a satisfying estimation level When no history of interactions with other agents in a new

environment Witness reputation: reports of witnesses about an agent’s

behavior, provided by other agents - Does not guarantee reliable estimation Are self-interested agents willing to share information? How much can you trust the informer?

IC-ININFO, Sep 19-22, 2016

Page 5: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 5

IMPLEMENTING TRUST / REPUTATION MODELS Centralized approach: One or more centralized trust authorities keep agent interaction

references (ratings) and give trust estimations Convenient for witness reputation models (e.g. eBay, SPORAS, etc.)

+Simpler to implement; better and faster trust estimations- Less reliable; Unrealistic: hard to enforce central controlling

authorities in open environments Decentralized (distributed) approach: Each agent keeps its own interaction references with other agents

and must estimate on its own the trust upon another agent Convenient for interaction trust models

+Robustness: no single point of failure; more realistic- Need more complex interaction protocols

IC-ININFO, Sep 19-22, 2016

Page 6: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 6

OTHER TRUST / REPUTATION MODELS Hybrid models: Combination of Interaction Trust and

Witness Reputation Regret / Social Regret FIRE RRAF / TRR CRM T-REX / HARM / DISARM

Certified reputation: third-party references provided by the agent itself Distributed approach for witness reputation

Centralized / Distributed

IC-ININFO, Sep 19-22, 2016

Page 7: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 7

SOME TERMINOLOGYWITNESS REPUTATION

Agent A wants a service from agent B Agent A asks agent C if agent B can be

trusted Agent C trusts agent B and replies yes

to A Agent A now trusts B and asks B to

perform the service on A’s behalf A = truster / beneficiary,

C = trustor / broker / consultant, B = trustee

IC-ININFO, Sep 19-22, 2016

ATruster /

Beneficiary

BTruste

e

CTrustor / Broker /

Consultant

?

?

Page 8: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 8

SOME TERMINOLOGYINTERACTION TRUST

Agent A wants a service from agent B Agent A judges if B is to be trusted

from personal experience Agent A trusts B and asks B to perform

the service on A’s behalf A = trustor / truster / beneficiary,

B = trustee

IC-ININFO, Sep 19-22, 2016

ATrustor / Truster /

Beneficiary

BTruste

e

?

Page 9: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 9

SOME TERMINOLOGYREFERENCES

Agent A wants a service from agent B Agent A asks agent B for proof of trust Agent B provides some agents R that

can guarantee that B can be trusted Agent A now trusts B and asks B to

perform the service on A’s behalf A = trustor / truster / beneficiary,

B = trustee,R = referee

IC-ININFO, Sep 19-22, 2016

ATruster /

Beneficiary

BTruste

e

RReferee

?

?

RReferee

Page 10: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents

10 CENTRALIZED TRUST MODELS

IC-ININFO, Sep 19-22, 2016

Page 11: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 11

EBAY-LIKE REPUTATION MODEL Centralised rating system After an interaction eBay users rate partner (−1, 0,+1) positive, neutral, negative

Ratings stored centrally Reputation value is computed as the sum of ratings over 6

months Reputation is a global single value

- Too simple for applications in MAS (1-D) All ratings count equally

- Cannot adapt well to changes in a user’s performance E.g. a user may cheat in a few interactions after obtaining a high

reputation value, but still retains a positive reputationIC-ININFO, Sep 19-22, 2016

Page 12: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 12

SPORASZacharia G. & Maes P.: Trust Management Through Reputation Mechanisms. Applied Artificial Intelligence, 14(9):881-908 (2000)

Centralized repository and reputation / reliability estimation Ratings are not kept – reputation updated after each

transaction New agents start with a minimum reputation

value and build up reputation The reputation value of an agent never falls

below the reputation of a newcomer Users with very high reputation values experience much smaller

rating changes after each update Ratings are discounted over time so that the most recent ratings

have more weight in the evaluation of an agent’s reputation+ More adaptable to changes in behavior over time

Reliability measure = ratings deviation High deviation: Agent has not been active enough to have a more accurate reputation

prediction Agent’s behaviour has a high degree of variation. IC-ININFO, Sep 19-22, 2016

+ Prevent agents with bad reputation leaving and entering with fresh reputation

- May discourage newcomers

Page 13: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents

13

DISTRIBUTED TRUST / REPUTATION MODELS

IC-ININFO, Sep 19-22, 2016

Page 14: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 14

REGRETSabater, J., & Sierra, C. (2001). Regret: A Reputation Model for Gregarious Societies. Proc. 4th Workshop On Deception, Fraud and Trust in Agent Societies, 61-69.

Decentralized reputation model Εach agent is able to evaluate the reputation of others by

itself Keeps own ratings about other partners in a local database Direct trust is calculated as the weighed average of all

ratings Each rating is weighed according to its recency More recent ratings are weighted more- Time granularity control does not actually reflect a rating’s recency

Reliability How many ratings are taken into account? Deviation of ratings

IC-ININFO, Sep 19-22, 2016

Page 15: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 15

SOCIAL REGRETSabater, J., & Sierra, C. (2002). Social Regret, A Reputation Model Based on Social Relations. Sigecom Exch, 3(1):44-56.

An attempt to locate witnesses’ ratings through a social graph Agent groups with frequent interactions among them Each group is a single source of reputation values Only the most representative agent within each group is asked for

information Heuristics are used to find groups and to select the best agent to

ask- Social Regret does not reflect the actual social relations

among agents Attempts to heuristically reduce the number of queries to be asked

in order to locate ratings- Considers the opinion of only one agent of each group Most agents are marginalized, distorting reality.

IC-ININFO, Sep 19-22, 2016

Page 16: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 16

REFERRAL SYSTEMYu, B., & Singh, M. P. (2002). An Evidential Model of Distributed Reputation Management. Proceedings 1st International Joint Conference on Autonomous Agents and Multi-agent Systems. Vol. 1. (pp. 294–301).

Decentralized, witness reputation model Agents cooperate by giving, pursuing, and evaluating referrals Each agent maintains a list of acquaintances (other agents it

knows) and their expertise When looking for certain information, an agent queries its

acquaintances They will try to answer the query, if possible If not, they will send back referrals of other agents they believe are

likely to have the desired information Agent’s expertise is used to determine how likely it is to have interaction with know witnesses of the target agent

IC-ININFO, Sep 19-22, 2016

Page 17: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 17

JURCA & FALTINGS REPUTATION MODELJurca, R. & Faltings, B.: Towards Incentive-compatible Reputation Management. In Trust, Reputation and Security: Theories and Practice, LNAI 2631, pp. 138–147, Springer (2003)

Agents are incentivized to report truthfully interactions’ results

Broker agents Buy and aggregate (average) reports from other agents Sell needed information to agents

+Mechanism guarantees that agents who report incorrectly will gradually lose money, while honest agents will not

±Brokers are distributed, but each one collects reputation values centrally

- Reputation values limited to 0, 1

IC-ININFO, Sep 19-22, 2016

Page 18: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 18

CERTIFIED REPUTATION (CR)Huynh, D, Jennings, N R., & Shadbolt, N R. (2006). Certified Reputation: How an Agent Can Trust a Stranger. In AAMAS '06: Proceedings 5th International Joint Conference on Autonomous Agents and Multiagent Systems, Hokkaido, Japan.

Decentralized witness reputation model Each agent keeps references given to it from other agents After every transaction, each agent asks partners to provide their

certified ratings about its performance It chooses the best ratings to store in its (local) database When an agent A contacts B to use service C, it asks B to provide

references about its past performance with respect to C Agent A receives certified ratings of B from B and calculates CR of

B

+Agents do not have costs involved in locating witness reports- Ratings might be misquoted Each agent only provides the best ratings about itself

IC-ININFO, Sep 19-22, 2016

Page 19: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents

19HYBRID TRUST / REPUTATION MODELS

IC-ININFO, Sep 19-22, 2016

Page 20: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 20

FIREHuynh, D., Jennings, N R., & Shadbolt, N. R. (2006). An Integrated Trust and Reputation Model for Open Multi-agent Systems. Journal of Autonomous Agents and Multi-agent Systems (AAMAS), 13(2):119-154.

Distributed, hybrid model Integrates 4 types of trust / reputation: interaction trust,

role-based trust, witness reputation and certified reputation Role-based trust: defined by various role-based relationships

between the agents Provides means for domain-specific customization of trust, using rules

All values are combined into a single measure by weighted average - Weak computation model for the combined reputation

estimation Does not take into consideration the problem of lying and

inaccuracy

IC-ININFO, Sep 19-22, 2016

Page 21: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 21

TRR (TRUST–RELIABILITY–REPUTATION)Rosaci, D., Sarne, G., Garruzzo, S. (2012). Integrating Trust Measures in Multiagent Systems. International Journal if Intelligent Systems, 27: 1–15.

Hybrid reputation model Combines interaction trust and witness reputation into a single

value, using weighted sum Dynamically computes weights Previous RRAF model used static weights assigned by the user

Weight depends on: Number of interactions between two agents Expertise that an agent has in evaluating other agents

capabilities Reputation is based on the global trust that community has

in an agent+More accurate reputation prediction- Difficult to implement in a distributed environment

IC-ININFO, Sep 19-22, 2016

Page 22: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 22

CRM (COMPREHENSIVE REPUTATION MODEL) Khosravifar, B., Bentahar, J., Gomrokchi, M., & Alam, R. (2012). CRM: An Efficient Trust and Reputation Model for Agent Computing. Knowledge-based Systems, 30:1-16.

Distributed, hybrid, probabilistic-based reputation model On-line trust estimation: Direct trust evaluation, using own interaction history Consulting reports: indirect trust estimation from consulting

agents trustworthy agents referee agents (introduced by the trustee agent as recommenders)

Off-line trust estimation: Off-line interaction inspection: After interaction, trustor assigns

a (useful/useless) flag for each consulting agent Rating & confidence of provided information Number and recency of interactions with the trustee agent

Maintenance: update information about consulting agents Initiated at different intervals of time

IC-ININFO, Sep 19-22, 2016

Page 23: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents

23

PRESENTER’S TRUST / REPUTATION MODELS

IC-ININFO, Sep 19-22, 2016

Page 24: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 24

PRESENTED TRUST / REPUTATION MODELS Centralized T-REX Hybrid

HARM Hybrid Knowledge-based Temporal Defeasible Logic

Distributed DISARM Hybrid Knowledge-based, Defeasible Logic Social relationships

IC-ININFO, Sep 19-22, 2016

Page 25: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents

25 T-REX

Kravari, K., Malliarakis, C., & Bassiliades, N.: T-REX: A Hybrid Agent Trust Model Based on Witness Reputation and Personal Experience. Proc. 11th International Conference on Electronic Commerce and Web Technologies (EC-Web 2010). Springer, LNBIP, Vol. 61, Part 3, pp. 107-118 (2010)

IC-ININFO, Sep 19-22, 2016

Page 26: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 26

T-REX OVERVIEW Centralized, hybrid reputation model Combines Witness Reputation (ratings) & Direct Trust

(personal experience) Weighted average (user-defined weights)

Ratings decay over time All past ratings are taken into account Time-stamps are used as weights Past ratings loose importance over time through a linear

extinguishing function+Low bandwidth, computational and storage cost

IC-ININFO, Sep 19-22, 2016

Page 27: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 27

Central ratings repository: Trustor A special agent responsible for collecting, storing, retrieving

ratings and calculating trust values Considered certified/reliable

Interacting agents Truster / Beneficiary: an agent that wants to interact with

another agent that offers a service Trustee: the agent that offers the service

Role of Trustor Before the interaction, Truster asks from Trustor calculation of

Trustees trust value After the interaction, Truster submits rating for Trustee’s

performance to Trustor

T-REX INTERACTION MODEL

IC-ININFO, Sep 19-22, 2016

Page 28: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 28IC-ININFO, Sep 19-22, 2016

Truster Trusteeinteract?

Trustor

T-REX RATING MECHANISM (I)

final reputation value

asks reputation

Calculates reputation from stored ratings and agent’s weights

Gives personalized weights for each rating criteria

Page 29: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 29

T-REX RATING MECHANISM (II)

IC-ININFO, Sep 19-22, 2016

TrusterTrusteeinteract

Trustor

Evaluation criteria• Correctness• Completeness• Response time

evaluation report

Page 30: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 30

Rating Vector

rx ∈ [0.1, 10] | 0.1 = terrible , 10 = perfect Normalized Ratings Ratings normalized logarithmically to

cross out extreme values log(rx) ∈ [-1, 1] | -1 = terrible, 1 = perfect

Weighted Normalized Ratings Agents customize the importance of

criteria using weights

T-REX RATING MECHANISM (III)

a: truster b: trustee Corr: correctness Resp: response

time Comp:

completeness t: time stamp px: weight of rating

rx

IC-ININFO, Sep 19-22, 2016

¿⃗ ¿

r ab (t )=|⃗Rab ( t )|=∑x=1

3

[ px ∙ log (r xab (t ) ) ]

∑x=1

3

px

Page 31: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 31

T-REX RATING MECHANISM (IV)The final reputation value

(TR)

Transactions between a and b

(interaction trust)

Transactions of b with all other agents

(witness report)

social trust weights (πp, πo)

Time is important more recent ratings “weigh”

more

𝑇𝑅𝑎𝑏 (𝑡 )=𝜋𝑝

𝜋𝑝+𝜋𝑜∙∑∀ 𝑡𝑖<𝑡

[𝑟𝑎𝑏 (𝑡𝑖 )∙ 𝑡 𝑖 ]

∑∀𝑡 𝑖< 𝑡

𝑡 𝑖+

𝜋𝑜

𝜋𝑝+𝜋𝑜∙ ∑∀ 𝑗≠ 𝑎 , 𝑗 ≠𝑏

∑∀𝑡 𝑖<𝑡

[𝑟 𝑗𝑏 ( 𝑡𝑖 ) ∙ 𝑡𝑖 ]

∑∀ 𝑡𝑖<𝑡

𝑡 𝑖

IC-ININFO, Sep 19-22, 2016

Page 32: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 32

T-REX ADVANTAGES Dynamicity Trust depends on time

Flexibility Can be customized from completely witness-based to

completely personal-experience-based Reliability Centralized repository (Trustor) is assumed honest

Low bandwidth Trusters report in one communication step Trustees do not communicate at all

Low storage cost Trusters and trustees do not store anything Trustor stores everything

Computational cost?IC-ININFO, Sep 19-22, 2016

Page 33: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 33

REDUCING COMPUTATIONAL COMPLEXITY (I) Original Formula

Computational complexity O(n*t) n: number of agents / t: amount of

transactions

Swap order of sums ti, x

Sums can be calculated incrementally, when a new report arrives

𝑇𝑅𝑎𝑏 (𝑡 )=𝜋𝑝

𝜋𝑝+𝜋𝑜∙

∑∀ 𝑡𝑖<𝑡 [∑𝑥=1

3

[𝑝 𝑥 ∙ log (𝑟 𝑥𝑎𝑏 (𝑡 ) ) ]

∑𝑥=1

3

𝑝 𝑥

∙ 𝑡𝑖]∑∀𝑡 𝑖< 𝑡

𝑡 𝑖+

𝜋𝑜

𝜋𝑝+𝜋𝑜∙ ∑∀ 𝑗≠ 𝑎 , 𝑗 ≠𝑏

∑∀𝑡𝑖<𝑡 [∑𝑥=1

3

[𝑝𝑥 ∙ log (𝑟𝑥𝑗𝑏 (𝑡 ) ) ]

∑𝑥=1

3

𝑝 𝑥

∙𝑡 𝑖]∑∀𝑡 𝑖< 𝑡

𝑡 𝑖

𝑇𝑅𝑎𝑏 (𝑡 )=𝜋𝑝

𝜋𝑝+𝜋𝑜∙∑𝑥=1

3

[𝑝𝑥 ∙∑∀ 𝑡𝑖<𝑡

log (𝑟 𝑥𝑎𝑏 (𝑡 ) ) ∙𝑡 𝑖]

∑𝑥=1

3

𝑝𝑥 ∙∑∀𝑡 𝑖<𝑡

𝑡 𝑖+

𝜋𝑜

𝜋𝑝+𝜋𝑜∙ ∑∀ 𝑗≠ 𝑎 , 𝑗≠ 𝑏

∑𝑥=1

3

[𝑝 𝑥 ∙∑∀𝑡 𝑖<𝑡

log (𝑟𝑥𝑗𝑏 (𝑡 ) )∙ 𝑡 𝑖]∑𝑥=1

3

𝑝𝑥 ∙ ∑∀𝑡𝑖<𝑡

𝑡𝑖

𝜎 𝑥𝑞𝑦 (𝑡 )=∑

∀𝑡 𝑖<𝑡log (𝑟𝑥𝑞𝑦 (𝑡 ) ) ∙ 𝑡𝑖 𝑇 (𝑡 )=∑

∀𝑡𝑖<𝑡𝑡𝑖

IC-ININFO, Sep 19-22, 2016

Page 34: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 34

REDUCING COMPUTATIONAL COMPLEXITY (II)

Computational complexity O(n) n: number of agents

Storage complexity O(n2) Swap order of sums j, x

This sum can be calculated incrementally

𝑇𝑅𝑎𝑏 (𝑡 )=𝜋𝑝

𝜋𝑝+𝜋𝑜∙∑

x=1

3

¿¿¿

𝑇𝑅𝑎𝑏 (𝑡 )=𝜋𝑝

𝜋𝑝+𝜋𝑜∙∑𝑥=1

3

¿¿¿

𝑆𝑥𝑎𝑏 (𝑡 )= ∑

∀ 𝑗≠𝑎 , 𝑗≠𝑏¿¿

IC-ININFO, Sep 19-22, 2016

Page 35: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 35

Computational complexity reduced to O(1)

Storage complexity remains O(n2)

REDUCING COMPUTATIONAL COMPLEXITY (III)

𝑇𝑅𝑎𝑏 (𝑡 )=𝜋𝑝

𝜋𝑝+𝜋𝑜∙∑𝑥=1

3

¿¿¿

IC-ININFO, Sep 19-22, 2016

Page 36: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents

36 HARM

Kravari, K., & Bassiliades, N. (2012). HARM: A Hybrid Rule-based Agent Reputation Model Based on Temporal Defeasible Logic. 6th International Symposium on Rules: Research Based and Industry Focused (RuleML-2012). Springer, LNCS 7438: 193-207.

IC-ININFO, Sep 19-22, 2016

Page 37: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 37

HARM OVERVIEW Centralized hybrid reputation model Combine Interaction Trust and Witness Reputation

Rule-based approach Temporal defeasible logic Non-monotonic reasoning

Ratings have a time offset Indicates when ratings become active to be considered for

trust assessment Intuitive method for assessing trust Related to traditional human reasoning

IC-ININFO, Sep 19-22, 2016

Page 38: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 38

(TEMPORAL) DEFEASIBLE LOGIC Temporal defeasible logic (TDL) is an extension of

defeasible logic (DL). DL is a kind of non-monotonic reasoning Why defeasible logic? Rule-based, deterministic (without disjunction) Enhanced representational capabilities Classical negation used in rule heads and bodies Negation-as-failure can be emulated Rules may support conflicting conclusions Skeptical: conflicting rules do not fire Priorities on rules resolve conflicts among rules Low computational complexity

IC-ININFO, Sep 19-22, 2016

Page 39: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 39

DEFEASIBLE LOGIC Facts: e.g. student(Sofia) Strict Rules: e.g. student(X) person(X) Defeasible Rules: e.g. r: person(X) works(X)

r’: student(X) ¬works(X) Priority Relation between rules, e.g. r’ > r Proof theory example: A literal q is defeasibly provable if: supported by a rule whose premises are all defeasibly provable

AND q is not definitely provable AND each attacking rule is non-applicable or defeated by a superior

counter-attacking rule

IC-ININFO, Sep 19-22, 2016

Page 40: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 40

TEMPORAL DEFEASIBLE LOGIC Temporal literals: Expiring temporal literals l:t

Literal l is valid for t time instances Persistent temporal literals l@t

Literal l is active after t time instances have passed and is valid thereafter Temporal rules: a1:t1 ... an:tn d b:tb d is the delay between the cause and the effect

Example:(r1) => a@1 Literal a is created due to r1. (r2) a@1=>7 b:3 It becomes active at time offset 1.

It causes the head of r2 to be fired at time 8.The result b lasts only until time 10.Thereafter, only the fact a remains.

IC-ININFO, Sep 19-22, 2016

Page 41: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 41

HARM – AGENT EVALUATED ABILITIES Validity An agent is valid if it is both sincere and credible Sincere: believes what it says Credible: what it believes is true in the world

Completeness An agent is complete if it is both cooperative and vigilant Cooperative: says what it believes Vigilant: believes what is true in the world

Correctness An agent is correct if its provided service is correct with

respect to a specification Response time Time that an agent needs to complete the transaction

IC-ININFO, Sep 19-22, 2016

Page 42: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 42

HARM - RATINGS Agent A establishes interaction with agent B: (A) Truster is the evaluating agent (B) Trustee is the evaluated agent

Truster’s rating value has 8 coefficients: 2 IDs: Truster, Trustee 4 abilities: Validity, Completeness, Correctness, Response time 2 weights (how much attention agent should pay on each rating?): Confidence: how confident the agent is for the rating Ratings of confident trusters are more likely to be right

Transaction value: how important the transaction was for the agent Trusters are more likely to report truthful ratings on important

transactions Example (defeasible RuleML / d-POSL syntax):

rating(id→1,truster→A,trustee→B,validity→5,completeness→6,correctness→6,resp_time→8,confidence→0.8,transaction_val→0.9).

IC-ININFO, Sep 19-22, 2016

Page 43: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 43

HARM – EXPERIENCE TYPES Direct Experience (PRAX ) Indirect Experience reports provided by strangers (SRAX) reports provided by known agents (e.g. friends) due to

previous interactions (KRAX ) Final reputation value of an agent X, required by an agent A

RAX = {PRAX , KRAX, SRAX}

IC-ININFO, Sep 19-22, 2016

Page 44: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 44

HARM – EXPERIENCE TYPES One or more rating categories may be missing E.g. a newcomer has no personal experience

A user is much more likely to believe statements from a trusted acquaintance than from a stranger. Personal opinion (AX) is more valuable than strangers’ opinion

(SX) and known partners (KX). Superiority relationships among rating categories

KX

AX, KX, SX

AX, KX

AX, SX

KX, SX

AX

SX

IC-ININFO, Sep 19-22, 2016

Page 45: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 45

HARM – FINAL REPUTATION VALUE RAX is a function that combines each available category personal opinion (AX) strangers’ opinion (SX) previously trusted partners (KX)

HARM allows agents to define weights of ratings’ coefficients Personal preferences

, ,AX AX AXAXR PR KR SR

4 4 4

1 1 1

log log log, , ,

, , , _

coefficient coefficient coefficienti AX i AX i AX

AX

i i ii i i

AVG w pr AVG w kr AVG w srR

w w w

coefficient validity completeness correctness response time

IC-ININFO, Sep 19-22, 2016

Page 46: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 46

r1: count_rating(rating→?idx, truster→?a, trustee→ ?x) := confidence_threshold(?conf), transaction_value_threshold(?

tran), rating(id→?idx, confidence→?confx, transaction_val→?tranx), ?confx >= ?conf, ?tranx >= ?tran.

r2: count_rating(…) := …?confx >= ?conf.

r3: count_rating(…) := …?tranx >= ?tran.

r1 > r2 > r3

HARM- WHICH RATINGS “COUNT”?

• if both confidence and transaction importance are high, then rating will be used for estimation

• if transaction value is lower than the threshold, but confidence is high, then use rating

• if there are only ratings with high transaction value, then they should be used

• In any other case, omit the rating

IC-ININFO, Sep 19-22, 2016

Page 47: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 47

HARM - CONFLICTING LITERALS All the previous rules are conclude positive literals. These literals are conflicting each other, for the same pair of

agents (truster and trustee) We want in the presence e.g. of personal experience to omit

strangers’ ratings. That’s why there is also a superiority relationship between the

rules. The conflict set is formally determined as follows:C[count_rating(truster→?a, trustee→?x)] =

{ ¬ count_rating(truster→?a, trustee→?x) } { count_rating(truster→?a1, trustee→?x1) | ?a ?a1

∧ ?x ?x1 }

IC-ININFO, Sep 19-22, 2016

Page 48: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 48

HARM - DETERMINING EXPERIENCE TYPES

known(agent1→?a, agent2→?y) :- count_rating(rating → ?id, truster→?a, trustee→?y).

count_pr(agent→?a, truster→?a, trustee→?x, rating→?id) :-count_rating(rating → ?id, truster→? a, trustee→ ?x).

count_kr(agent→?a, truster→?k, trustee→?x, rating →?id) :-known(agent1→?a, agent2→?k), count_rating(rating→?id, truster→?k, trustee→ ?x).

count_sr(agent→?a, truster→?s, trustee→?x, rating→?id) :-count_rating(rating → ?id, truster →?s, trustee→ ?x), not(known(agent1→?a, agent2→?s)).

Which agents are considered as known?

Rating categories

IC-ININFO, Sep 19-22, 2016

Page 49: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 49

HARM – SELECTING EXPERIENCES

Final step is to decide whose experience will “count”: direct, indirect (witness), or both.

The decision for RAX is based on a relationship theory e.g. Theory #1: All categories count equally.r8: participate(agent→?a, trustee→?x, rating→?id_ratingAX) :=

count_pr(agent→?a, trustee→?x, rating→ ?id_ratingAX).r9: participate(agent→?a, trustee→?x, rating→?id_ratingKX) :=

count_kr(agent→?a, trustee→?x, rating→ ?id_ratingKX).r10: participate(agent→?a, trustee→?x, rating→?id_ratingSX) :=

count_sr(agent→?a, trustee→?x, rating→ ?id_ratingSX).

IC-ININFO, Sep 19-22, 2016

Page 50: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 50

KX

AX, KX, SX

AX, KX

AX, SX

KX, SX

AX

SX

SELECTING EXPERIENCES ALL CATEGORIES COUNT EQUALLY

IC-ININFO, Sep 19-22, 2016

Page 51: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 51

SELECTING EXPERIENCES - THEORY #2 PERSONAL EXPERIENCE IS PREFERRED TO FRIENDS’ OPINION TO STRANGERS’ OPINIONr8: participate(agent→?a, trustee→?x, rating→?id_ratingAX) :=

count_pr(agent→?a, trustee→?x, rating→ ?id_ratingAX).r9: participate(agent→?a, trustee→?x, rating→?id_ratingKX) :=

count_kr(agent→?a, trustee→?x, rating→ ?id_ratingKX).r10: participate(agent→?a, trustee→?x, rating→?id_ratingSX) :=

count_sr(agent→?a, trustee→?x, rating→ ?id_ratingSX).r8 > r9 > r10

IC-ININFO, Sep 19-22, 2016

Page 52: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 52

KX

AX, KX, SX

AX, KX

AX, SX

KX, SX

AX

SX

SELECTING EXPERIENCES PERSONAL EXPERIENCE IS PREFERRED TO FRIENDS’ OPINION TO STRANGERS’ OPINION

>>

IC-ININFO, Sep 19-22, 2016

Page 53: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 53

SELECTING EXPERIENCES - THEORY #3PERSONAL EXPERIENCE AND FRIENDS’ OPINION IS PREFERRED TO STRANGERS’ OPINIONr8: participate(agent→?a, trustee→?x, rating→?id_ratingAX) :=

count_pr(agent→?a, trustee→?x, rating→ ?id_ratingAX).r9: participate(agent→?a, trustee→?x, rating→?id_ratingKX) :=

count_kr(agent→?a, trustee→?x, rating→ ?id_ratingKX).r10: participate(agent→?a, trustee→?x, rating→?id_ratingSX) :=

count_sr(agent→?a, trustee→?x, rating→ ?id_ratingSX).r8 > r10, r9 > r10

IC-ININFO, Sep 19-22, 2016

Page 54: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 54

KX

AX, KX, SX

AX, KX

AX, SX

KX, SX

AX

SX

SELECTING EXPERIENCESPERSONAL EXPERIENCE AND FRIENDS’ OPINION IS PREFERRED TO STRANGERS’ OPINION

>IC-ININFO, Sep 19-22, 2016

Page 55: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 55

HARM - TEMPORAL DEFEASIBLE LOGIC EXTENSION Agents may change their behavior / objectives at any time Evolution of trust over time should be taken into account Only the latest ratings participate in the reputation estimation

In the temporal extension of HARM: each rating is a persistent temporal literal of TDL each rule conclusion is an expiring temporal literal of TDL

Truster’s rating is active after time_offset time instances have passed and is valid thereafter

rating(id→val1, truster→val2, trustee→ val3, validity→val4, completeness→val5, correctness→val6, resp_time→val7, confidence→val8, transaction_val→value9)@time_offset.

IC-ININFO, Sep 19-22, 2016

Page 56: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 56

HARM - TEMPORAL DEFEASIBLE LOGIC EXTENSION

Rules are modified accordingly: each rating is active after t time instances have passed each conclusion has a duration that it holds each rule has a delay between the cause and the effect

count_rating(rating→?idx, truster→?a, trustee→?x):duration :=delay

confidence_threshold(?conf),transaction_value_threshold(?tran), rating(id→?idx, confidence→?

confx,transaction_value→?tranx)@t, ?confx >= ?conf, ?tranx >= ?tran.

IC-ININFO, Sep 19-22, 2016

Page 57: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents

57 DISARM

K. Kravari, N. Bassiliades, “DISARM: A Social Distributed Agent Reputation Model based on Defeasible Logic”, Journal of Systems and Software, Vol. 117, pp. 130–152, July 2016

IC-ININFO, Sep 19-22, 2016

Page 58: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 58

DISARM OVERVIEW Distributed extension of HARM Distributed hybrid reputation model Combines Interaction Trust and Witness Reputation Ratings are located through agent’s social relationships

Rule-based approach Defeasible logic Non-monotonic reasoning

Time is directly used in: Decision making rules about recency of ratings Calculation of reputation estimation (similar to T-REX)

Intuitive method for assessing trust Related to traditional human reasoning

IC-ININFO, Sep 19-22, 2016

Page 59: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 59

DISARM - SOCIAL RELATIONSHIPS Social relationships of trust among agents If an agent is satisfied with a partner it is more likely to

interact again in the future If dissatisfied it will not interact again

Each agent maintains 2 relationship lists: White-list: Trusted agents Black-list: Non-trusted agents All other agents are indifferent (neutral zone)

Each agent decides which agents are added / removed from each list, using rules

Personal social network

IC-ININFO, Sep 19-22, 2016

Page 60: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 60

Truster’s rating value has 11 coefficients: 2 IDs: Truster, Trustee 4 abilities: Validity, Completeness, Correctness, Response

time 2 weights: Confidence, Transaction value Timestamp Cooperation: willingness to do what is asked for Important in distributed social environments

Outcome feeling: (dis)satisfaction for the transaction outcome Degree of request fulfillment

Example (defeasible RuleML / d-POSL syntax):

DISARM - RATINGS

IC-ININFO, Sep 19-22, 2016

rating (id→1, truster→A, trustee→X, t→140630105632, resp_time→9,

validity→7, completeness→6, correctness→6, cooperation→8, outcome_feeling→7, confidence→0.9, transaction_val→0.8)

3 more than HARM

Page 61: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 61

DISARM MODEL

IC-ININFO, Sep 19-22, 2016

...2. Propagating request

Agent ATruster

WL agents providing ratings1. Request ratings

3. Receive ratings

Agent XTrustee

5. Choose agent xService request

6. Receive service

4. Evaluate Reputation(DISARM rules + ratings)

Page 62: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 62

DISARM - BEHAVIOR CHARACTERIZATION

good_behavior(time → ?t, truster→ ?a, trustee→ ?x, reason → all) :- resp_time_thrshld(?resp), valid_thrshld(?val), …,

trans_val_thrshld(?trval),rating(id→?idx, time → ?t, truster→ ?a, trustee→ ?x, resp_time→?

respx, validity→?valx, transaction_val→?trvalx, completeness→?

comx, correctness→?corx, cooperation→?coopx,

outcome_feeling→?outfx), ?respx<?resp, ?valx>?val, ?comx>?com, ?corx>?cor, ?coopx>?coop,

?outfx>?outf.bad_behavior(time → ?t, truster→ ?a, trustee→ ?x, reason → response_time) :-

rating(id→?idx, time → ?t, truster→ ?a, trustee→ ?x, resp_time→?respx),

resp_time_thrshld(?resp), ?respx >?resp.

Any combination of parameters can be used with any defeasible theory.

IC-ININFO, Sep 19-22, 2016

Page 63: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 63

DISARM - DECIDING WHO TO TRUST Has been good twice for the same reasonadd_whitelist(trustee→ ?x, time → ?t2) :=

good_behavior(time→?t1, truster→?self, trustee→?x, reason→?r),

good_behavior(time→?t2, truster→?self, trustee→?x, reason→?r),

?t2 > ?t1. Has been bad thrice for the same reasonadd_blacklist(trustee→ ?x, time → ?t3) :=

bad_behavior(time→?t1, truster→?self, trustee→?x, reason→?r),bad_behavior(time→?t2, truster→?self, trustee→?x, reason→?r),bad_behavior(time→?t3, truster→?self, trustee→?x, reason→?r),?t2 > ?t1, ?t3 > ?t2.

IC-ININFO, Sep 19-22, 2016

Page 64: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 64

DISARM - MAINTAINING RELATIONSHIP LISTS

blacklist(trustee→ ?x, time → ?t) :=¬whitelist(trustee→ ?x, time → ?t1), add_blacklist(trustee→ ?x, time → ?t2), ?t2 > ?t1.

¬blacklist(trustee→ ?x, time → ?t2) :=blacklist(trustee→ ?x, time → ?t1),add_whitelist(trustee→ ?x, time → ?t2),?t2 > ?t1.

whitelist(trustee→ ?x, time → ?t) :=¬blacklist(trustee→ ?x, time → ?t1), add_whitelist(trustee→ ?x, time → ?t2), ?t2 > ?

t1.…

IC-ININFO, Sep 19-22, 2016

Add to the blacklist

Remove from the blacklist

Add to the whitelist

Page 65: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 65

DISARM - LOCATING RATINGS Ask for ratings about an agent sending request messages To whom and how? To everybody To direct “neighbors” of the agent’s “social network” To indirect “neighbors” of the “social network” though

message propagation for a predefined number of hops (Time-to-Live - P2P)

“Neighbors” are the agents in the whitelist Original request:send_message(sender→?self, receiver→?r,

msg →request_reputation(about→?x,ttl→?t)) :=ttl_limit(?t), whitelist(?r), locate_ratings(about→?

x).IC-ININFO, Sep 19-22, 2016

Page 66: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 66

DISARM - HANDLING RATINGS REQUEST Upon receiving request, return rating to the sendersend_message(sender→?self, receiver→?s,

msg →rating(id→ idx, truster→ ?self, trustee→ ?x, …)) :=

receive_message(sender→?s, receiver→?self, msg →request_rating(about→?x)),

rating(id→?idx, truster→ ?self, trustee→ ?x, …). If time-to-live has not expired propagate request to all friendssend_message(sender→?s, receiver→?r,

msg →request_reputation(about→?x, ttl→?t1)):=receive_message(sender→?s, receiver→?self,

msg →request_rating(about→?x,ttl→?t)), ?t >0, WL(?r), ?t1 is ?t - 1.

IC-ININFO, Sep 19-22, 2016

Page 67: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 67

DISARM - RATING CATEGORIES Direct Experience (PRX) Indirect Experience (reports provided by other agents): “Friends” (WRX) – agents in the whitelist Known agents from previous interactions (KRX) Complete strangers (SRX)

Final reputation value RX = {PRX, WRX, KRX, SRX}

IC-ININFO, Sep 19-22, 2016

KR

PR, WR, KR, SR

PR, WR PR, SR KR, SR

PR SR

PR, KR

WR

WR, SRWR, KR

PR, WR, SR WR, KR, SRPR, KR, SRPR, WR, KR

New compared to HARM

Page 68: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 68

DISARM - SELECTING RATINGS

According to user’s preferenceseligible_rating(rating→?idx,truster→?a,trustee→?x,reason→cnf_imp) :=

conf_thrshld(?conf), trans_val_thrshld(?tr),rating(id→?idx,truster→?a,trustee→?x,conf→?confx,trans_val→?

trx),?confx >= ?conf, ?trx >= ?tr.

According to temporal restrictionscount_rating(rating→?idx, truster→?a, trustee→?x) := time_from_thrshld(?ftime), time_to_thrshld(?ttime), rating(id→?idx, t→?tx, truster→?a, trustee→ ?x),

?ftime <=?tx <= ?ttime.IC-ININFO, Sep 19-22, 2016

Page 69: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 69

DISARM – DETERMINING RATING CATEGORIES

count_wr (rating →?idx, trustee→?x) :-eligible_rating(rating → ?idx, cat→?c, truster→?k, trustee→ ?x), count_rating(rating→?idx, truster→?k, trustee→ ?x),known(agent→?k), whitelist (trustee →?k).

count_kr (rating →?idx, trustee→?x) :-eligible_rating(rating→?idx, cat→?c, truster→?k, trustee→ ?x),count_rating(rating→?idx, truster→?k, trustee→ ?x),known(agent→?k), not(whitelist(trustee →?k)),not(blacklist (trustee →?k)).

IC-ININFO, Sep 19-22, 2016

Page 70: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 70

DISARM - FACING DISHONESTY When ratings provided by an agent are outside the standard

deviation of all received ratings, the agent might behave dishonestly

bad_assessment (time → ?t, truster→ ?y, trustee→ ?x) :- standard_deviation_value(?t,?y,?x,?stdevy), standard_deviation_value (?t,_,?x,?stdev), ?stdevy > ?stdev.

When two bad assessments for the same agent were given in a certain time window, trust is lost

remove_whitelist(agent→ ?y, time → ?t2) :=whitelist(truster→ ?y),time_window(?wtime), bad_assessment(time → ?t1, truster→ ?y, trustee→ ?x),bad_assessment(time → ?t2, truster→ ?y, trustee→ ?x),?t2 <= ?t1 + ?wtime.

IC-ININFO, Sep 19-22, 2016

Page 71: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents

71

EVALUATION OF TRUST / REPUTATION MODELS

IC-ININFO, Sep 19-22, 2016

Page 72: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 72

EVALUATION ENVIRONMENT Simulation in the EMERALD* multi-agent

system Service provider agents All provide the same service

Service consumer agents Choose provider with the higher reputation

value Performance metric: Utility Gain

*K. Kravari, E. Kontopoulos, N. Bassiliades, “EMERALD: A Multi-Agent System for Knowledge-based Reasoning Interoperability in the Semantic Web”, 6th Hellenic Conference on Artificial Intelligence (SETN 2010), Springer, LNCS 6040, pp. 173-182, 2010.

IC-ININFO, Sep 19-22, 2016

Number of simulations: 500

Number of providers: 100

Good providers 10

Ordinary providers 40

Intermittent providers 5

Bad providers 45

Page 73: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 73

T-REX PERFORMANCE OVER TIME

IC-ININFO, Sep 19-22, 2016

Page 74: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 74

HARM VS. T-REX VS. STATE-OF-THE-ART

IC-ININFO, Sep 19-22, 2016

HARM T-REX No Trust CR SPORAS5.73 5.57 0.16 5.48 4.65

Page 75: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 75

Series10

1

2

3

4

5

6

7

DISARM Social Regret Certified Reputation CRM FIREHARM NONE

Time

Mea

n UG

DISARM VS. HARM VS. STATE-OF-THE-ART

IC-ININFO, Sep 19-22, 2016

Mean Utility Gain

Page 76: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 76

DISARM - ALONE

IC-ININFO, Sep 19-22, 2016

Better performance when alone, due to more social relationships

Page 77: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 77

Series10

5

10

15

20

25

30

DISARM Social Regret Certified Reputation CRM FIRE

Time

Mem

ory

Spac

ec (%

)DISARM VS. HARM VS. STATE-OF-THE-ART

IC-ININFO, Sep 19-22, 2016

Storage Space

Page 78: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 78

EVALUATING DISHONESTY HANDLING

IC-ININFO, Sep 19-22, 2016

Page 79: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents

79 CONCLUDING…

IC-ININFO, Sep 19-22, 2016

Page 80: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 80

TRUST / REPUTATION MODELS FOR MULTIAGENT SYSTEMS

Interaction Trust (personal experience) vs. Witness Reputation (Experience of others) Hybrid models

Centralized (easy to locate ratings) vs. Distributed (more robust)

Several State-of-the-Art models have been presented SPORAS, Regret, Certified Reputation, Referral, FIRE, TRR, CRM,

… Presenter’s and associates trust / reputation models T-REX (centralized, hybrid, time decay, computationally

optimized) HARM (centralized, hybrid, knowledge-based, temporal

defeasible logic) DISARM (distributed, hybrid, knowledge-based, defeasible logic,

time decay, social relationships, manages dishonesty)IC-ININFO, Sep 19-22, 2016

Page 81: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 81

CONCLUSIONS Centralized models

+Achieve higher performance because they have access to more information

+Simple interaction protocols, easy to locate ratings+Both interaction trust and witness reputation can be easily

implemented- Single-point-of-failure- Cannot scale well (bottleneck, storage & computational complexity)- Central authority hard to enforce in open multiagent systems

Distributed models- Less accurate trust predictions, due to limited information- Complex interaction protocols, difficult to locate ratings- More appropriate for interaction trust +Robust – no single-point-of-failure+Can scale well (no bottlenecks, less complexity)+More realistic in open multiagent systems

IC-ININFO, Sep 19-22, 2016

Page 82: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 82

CONCLUSIONS Interaction trust

+More trustful- Requires a long time to reach a satisfying estimation level

Witness reputation- Does not guarantee reliable estimation+Estimation is available from the beginning of entering a

community Hybrid models

+Combine interaction trust and witness reputation- Combined trust metrics are usually only based on arbitrary /

experimentally-optimized weights

IC-ININFO, Sep 19-22, 2016

Page 83: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 83

Centralized models- Cannot scale well (bottleneck, storage & computational complexity)+T-REX reduces computational complexity at the expense of storage

complexity+HARM reduces computational complexity by reducing considered ratings,

through rating selection based on user’s domain-specific knowledge Distributed models

- Less accurate trust predictions, due to limited information- Complex interaction protocols, difficult to locate ratings+DISARM finds ratings through agent social relationships and increases

accuracy by using only known-to-be-trustful agents Hybrid models

- Combined trust metrics are usually only based on arbitrary weights+HARM & DISARM employ a knowledge-based highly-customizable (both

to user prefs & time) approach, using non-monotonic defeasible reasoning

CONCLUSIONS – PRESENTED MODELS

IC-ININFO, Sep 19-22, 2016

Page 84: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 84

A FEW WORDS ABOUT US… Aristotle University of Thessaloniki, Greece Largest University in Greece and South-East Europe Since 1925, 41 Departments, ~2K faculty, ~80K students

Dept. of Informatics Since 1992, 28 faculty, 5 research labs, ~1100 undergraduate

students, ~200 MSc students, ~80 PhD students, ~120 PhD graduates, >3500 pubs

Software Engineering, Web and Intelligent Systems Lab 7 faculty, 20 PhD students, 9 Post-doctorate affiliates

Intelligent Systems group (http://intelligence.csd.auth.gr) 4 faculty, 9 PhD students, 16 PhD graduates Research on Artificial Intelligence, Machine Learning / Data Mining,

Knowledge Representation & Reasoning / Semantic Web, Planning, Multi-Agent Systems

425 publications, 35 projectsIC-ININFO, Sep 19-22, 2016

Page 85: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 85

ACKNOWLEDGMENTS The work described in this talk has been performed in

cooperation with Dr. Kalliopi Kravari Former PhD student, currently postdoctorate affiliate

Occasional contributors: Dr. Christos Malliarakis (former MSc student, co-author) Dr. Efstratios Kontopoulos (former PhD student, co-author) Dr. Antonios Bikakis (Lecturer, University College London, PhD

examiner)

IC-ININFO, Sep 19-22, 2016

Page 86: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 86

RELEVANT PUBLICATIONS K. Kravari, C. Malliarakis, N. Bassiliades, “T-REX: A hybrid agent trust

model based on witness reputation and personal experience”, Proc. 11th International Conference on Electronic Commerce and Web Technologies (EC-Web 2010), Bilbao, Spain, Lecture Notes in Business Information Processing, Vol. 61, Part 3, Springer, pp. 107-118, 2010.

K. Kravari, E. Kontopoulos, N. Bassiliades, “EMERALD: A Multi-Agent System for Knowledge-based Reasoning Interoperability in the Semantic Web”, 6th Hellenic Conference on Artificial Intelligence (SETN 2010), Springer, LNCS 6040, pp. 173-182, 2010.

K. Kravari, N. Bassiliades, “HARM: A Hybrid Rule-based Agent Reputation Model based on Temporal Defeasible Logic”, 6th International Symposium on Rules: Research Based and Industry Focused (RuleML-2012). Springer Berlin/Heidelberg, LNCS, Vol. 7438, pp. 193-207, 2012.

K. Kravari, N. Bassiliades, “DISARM: A Social Distributed Agent Reputation Model based on Defeasible Logic”, Journal of Systems and Software, Vol. 117, pp. 130–152, July 2016

IC-ININFO, Sep 19-22, 2016

Page 87: Trust and reputation models among agents

N. Bassiliades - Trust and reputation models among agents 87IC-ININFO, Sep 19-22, 2016