Trust-based Service Composition and Binding with Multiple Objective Optimization in Service- Oriented Mobile Ad Hoc Networks Yating Wang†, Ing-Ray Chen†, Jin-Hee Cho*, Ananthram Swami* and Kevin S. Chan*

Introduction Service-oriented mobile ad hoc network (MANET) is populated with service providers (SPs) and service requesters (SRs) In this paper, the authors are concerned with: satisfying user service requests with multiple objectives including maximizing quality-of-service (QoS) and quality-of-information (QoI) while minimizing the service cost with user satisfaction (US) ultimately measuring success multi-objective optimization (MOO).

SYSTEM MODEL Two roles: service provider (SP) and service requestor (SR) Example: A user in a smart city issues a service request “take me to a nice Thai restaurant nearby with drunken noodle on its menu” with a service quality specified in terms of QoI, QoS, and cost for the overall service Service composition phase: transportation + food Service binding phase: select the best SPs out of all SPs available to the user at the time the service request is issued

SYSTEM MODEL Service Quality Criteria: QoI, service Delay (as a QoS attribute), and Cost. Normalization: Max and Min are known a priori Q,D,C MOO => multi-objective maximization

SYSTEM MODEL Malicious behaviors: Self-promotion: reporting false service quality information Opportunistic service: “just enough” service Bad-mouthing attack (BMA): providing bad recommendations Ballot stuffing attack (BSA): boost the reputation for bad nodes Packet dropping: drop packets

SERVICE COMPOSITION AND BINDING Service Advertisement Reply

SERVICE COMPOSITION AND BINDING Service composition specification (SCS): Constraints:

SERVICE COMPOSITION AND BINDING Service Binding SP can only participate in one service request at a time to ensure its availability and commitment to a single service request.

PROBLEM DEFINITION AND METRICS parallel structure series structure

PROBLEM DEFINITION AND METRICS MOO Problem Formulation system level: Preferences of SR

PROBLEM DEFINITION AND METRICS User Satisfaction: different from MOO value ratio of the actual service quality received to the best service quality available among SPs for executing Om Compare with USTm: user satisfaction threshold

TRUST MANAGEMENT PROTOCOL Trust management schemes: BRS: single-trust beta reputation system Multi-trust Protocol Design threshold-based relationship model (TRM) scaling relationship model (SRM)

TRUST MANAGEMENT PROTOCOL Single-trust Baseline Protocol Design (BRS) A positive evidence is observed when SRm is satisfied (USm exceeds USTm )

TRUST MANAGEMENT PROTOCOL Multi-trust Protocol Design Competence: intrinsic service capability, “true” Q, D, and C scores Integrity: degree complies with the protocols

TRUST MANAGEMENT PROTOCOL Multi-trust Protocol Design Still How to count ? For competence: the same with BRS For integrity: positive if SR sees node j’s observed Q, D and C scores are close to node j’s advertised scaled Q, D, and C scores Compare with

TRUST MANAGEMENT PROTOCOL Trust formation Threshold-based relationship model (TRM) : Scaling relationship model (SRM): More strict

ALGORITHM DESCRIPTION Four algorithms Non-trust-based BRS TRM SRM

ALGORITHM DESCRIPTION Non-trust-based blacklist of SPs: randomly select

ALGORITHM DESCRIPTION Trust-based Modify By multiplying to each single node in the bottom layer

RESULTS AND ANALYSIS Experiment Setup Proposed method: Heuristic-based solution (linear runtime complexity) SR ranks all eligible SPs for executing an abstract service and selects the highest ranked SP as the winner for executing that particular abstract service Optimal solution to be compared: Integer Linear Programming (exponential runtime complexity )

RESULTS AND ANALYSIS Comparative Performance Analysis

RESULTS AND ANALYSIS Effect of Service Quality Constraints and Opportunistic Service Attacks

RESULTS AND ANALYSIS Effect of Q, D, C Score Distribution