Tulungan: A slandering-resistant reputation system for collaborative web filtering systems

• Main Authors: Pantola, Alexis V., Festin, Susan Pancho, Salvador, Florante R.
• Format: text
• Published: Animo Repository 2013
• Subjects: Online reputation management; Information filtering systems; Computer Sciences
• Online Access: https://animorepository.dlsu.edu.ph/faculty_research/9256
• Institution: De La Salle University

Reputation systems measure the credibility of contributions and contributors in collaborative web systems. Measuring the credibility is significant since a collaborative environment generally allows anyone with Internet access to provide contribution. Collaborative web systems are susceptible to malicious users who intentionally provide inaccurate contents. With the help of reputation systems, the effect of such malicious activities can be reduced. Reputation systems allow Internet users to rate the contributions made by other users. However, there are malicious users who will go beyond providing wrong contributions. They attempt to make reputation systems useless by launching attacks such as slandering. Slandering happens when a malicious user or group of malicious users intentionally provide a negative rating to accurate contributions provided by good users. Such activity lowers the reputation of good users and in most cases it even helps improve the reputation of slandering users. This paper presents a reputation system called Tulungan that is designed to measure the contributor and rater reputation of users of a collaborative web system that is used for web filtering. User contributions are in the form of URL categorizations. It is the role of Tulungan to determine the correctness of the categorizations. A simulation is presented to validate the resilience of Tulungan in the presence of slandering users. The result of the simulation shows that Tulungan is not only resistant to slandering but it is still effective even if the number of good users is less than its slandering counterpart. Even if there are only 15% good users, the number of correct URL categorizations outnumbers incorrect contributions.