International Research Journal of Engineering and Technology (IRJET) Volume: 04 Issue: 02 | Feb -2017
www.irjet.net
e-ISSN: 2395 -0056 p-ISSN: 2395-0072
A SURVEY ON VARIOUS REPUTATION ASSESSMENT TECHNIQUES Sophia Kuriakose1 1PG
student, Dept. of Computer Engineering, VJCET, Vazhakulam, Kerala, India
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract – Nowadays evaluating the reputation of products has become an important key factor. Reputation is a result of social evaluation that helps to determine the level of trustworthiness on a set of criteria. In general, it is defined as a component of identity. Trustworthiness provides an environment which enables the users to interact each other with much ease. Trust and reputation plays an important role in almost all social platforms. This paper focuses on various reputation evaluation techniques and determines how trustworthiness could be attained in a more specific way. Key Words: Reputation, trust, malicious ratings, PHAT, Belief propagation
1. INTRODUCTION Almost in all online based purchases, consumers share their opinion regarding the products through ratings. The overall reputation of the products is analyzed by taking the aggregated score of all ratings given by the users. Reputation plays an important role in evaluating the products. Trustworthiness of a reputation is yet another factor to be considered as it can be manipulated easily by the occurrence of malicious ratings. Various algorithms and techniques are developed in order to achieve trustworthiness. Some of them involve using clustering, classification, probability distributions, and hypothesis test methods etc for evaluating reputation and to eliminate malicious ratings, thereafter make the system trustworthy. As various reputation measurement techniques are still at the infancy level, there are still many uninvestigated topics and areas worth research attention in order to fully leverage the potential of trustworthiness. In this paper different methods for evaluating reputation and thereby eliminating the occurrence of malicious ratings are studied so as to achieve trustworthiness.
2. LITERATURE SURVEY In [1] Hyun-Kyo Oh, Sang-Wook Kim, Sunju Park, and Ming Zhou proposed a true reputation algorithm where the reputation was estimated depending on the confidence of the ratings given by the customers. The framework was designed so as to mitigate the occurrence of malicious ratings and thereby evaluate the trustworthiness of reputation. © 2017, IRJET
|
Impact Factor value: 5.181
|
Confidence of rating was estimated based on user’s activity, objectivity and consistency. User’s activity was determined by the number of ratings given by the customers, objectivity was analyzed based on the deviation of rating and reputation and consistency was analyzed using box plot. Reputation of the products was adjusted based on the confidence of the rating. In [2] Yafei Yang, Yan (Lindsay) Sun, Steven Kay, and Qing Yang proposed a set of statistical methods for the purpose of detecting collaborative malicious ratings and thereby a framework of trust enhanced rating aggregation system was developed. The major components involved in trust enhanced rating aggregation system are: Rating aggregator and Trust manager. Rating aggregation process was analyzed using arrival rate detection, model change detection, histogram detection and mean change detection which were applied independently. Trust manager analyzes the output which in turn helps to evaluate the trustworthiness of individual raters. The malicious ratings were removed rating filter and thereby the ratings are combined by the rating aggregation algorithm. In [3] Xiaofeng Wang, Ling Liu, and Jinshu Su proposed a RLM model for the robust and diverse evaluation of reputation. Reputation evaluation method was determined by the two attributes such as reputation value and reputation prediction variance. Aggregation of feedbacks was computed using reputation prediction variance. Kalman aggregation method was designed for providing a robust evaluation of trust and also to aggregate feedbacks more accurately. Expectation Maximization algorithm was designed so as to defend against malicious and coordinated feedbacks and to obtain fair ratings. In addition to this a hypothesis testing is conducted to resist against malicious feedbacks from occurring. In [4] Noura Limam and Raouf Boutaba presents a framework for reputation aware software service selection and rating. Selection algorithm is introduced in order to provide service recommendation with the aim of providing best possible choices for SaaS consumers. The best possible choices rely on certain factors like quality, cost and trust. Rating function is derived for the purpose of monitoring results and service cost and thereby to produce feedbacks without human intervention. Finally a reputation derivation model aggregates all the feedbacks into reputation considering time factor as a parameter and as a result the service ranking function aggregates the parameters such as ISO 9001:2008 Certified Journal
|
Page 1007