Secure Personalized Recommendation System for Mobile User - Information Security and Cryptology-ICISC 2010 - page 266

Secure Personalized Recommendation System for Mobile
User
Soe Yu Maw
Computer University, Myeiktilar, Myanmar
[e-mail]
Abstract. Nowadays, due to the rapid growth of the mobile users,
personalization and recommender systems have gained popularity. The
recommender systems serve the personalized information to the users according
to user preferences or interests and their profiles. Tourism is an industry which
had adopted the use of new technologies. Recently, mobile tourism has come
into spotlight. Due to the rapid growing of user needs in mobile tourism
domain, we concentrated on to gives the personalized recommendation based
on multi-agent technology in tourism domain to serve the mobile users [7]. The
objective of this paper is to build a secure personalized recommendation
system. Attackers can affect the prediction of the recommender system by
injecting a number of biased profiles. In this paper, we consider detecting or
preventing the profile injection (also called shilling attacks) by using significant
weighting and trust weighting that complements to our proposed RPCF
Algorithm.
Keywords: Security, Personalization, Recommender System, Collaborative
Filtering, Profile Injection Attacks, RPCF Algorithm, Significant Weighting,
Trust Weighting.
1 Introduction
The ever-changing trends of our lifestyle require mobility supports [15] which open
up new accessibility opportunities for tourism industry. Today, tourism systems are
one of the most important application areas for recommender system. To cope with
the demand for quality of access, tourism information system should be made
ubiquitous, time-aware, location-aware and personalized.
In Modern world, personalization and recommendation systems have gained wide-
spread acceptance and attracted increased public interest in commercial services [6].
Collaborative filtering (CF) provides personalized recommendations, based on
suggestions of users with similar preferences. The development of CF algorithms has
focused mainly on how to provide accurate recommendations.
Recommender systems based on CF have the issues for the process of finding
similar users. An attacker can attempt to influence the behavior of the recommender
system for other users by using biased (fake) rating profiles to artificially either
promote or demote a target item. Such attacks have been referred to as shilling attacks
or profile injection attacks, and attackers as shillers [5]. Since user profiles of shillers