An Interaction-Based and Graph-Based Hybrid Approach to Evaluate Trust in Online Social Networks (OSNs)

With the digital revolution and the Web 2.0 era, web-based social networks such as Facebook and others have become popular mediums for users to do various activities. While social networks are becoming increasingly popular, concerns about trust and trust-related issues are also growing among users. There are many applications where trust plays a vital role in users’ decision-making, requiring trust evaluation. There are several trust evaluation approaches for online social networks in the literature. However, the existing approaches focus only on certain aspects and believe that direct trust between participants is known. Thus, there is a need for a comprehensive trust evaluation approach that infers indirect trust and strives to measure direct trust. This paper proposes an interaction-based and graph-based hybrid approach that attempts to measure direct trust and infer indirect trust among users. Our direct trust measure method utilizes the most important features and similarities between users to measure direct trust. The proposed indirect trust inference method uses the graph theory concept to infer indirect trust. We implement the friend-request identification and the Sybil attack detection applications using the proposed direct trust measure method. Both the applications are evaluated on synthetic and real-world datasets. The empirical results show that the friend-request identification application achieves a high accuracy of 96.17%, and the Sybil attack detection application obtains a high detection rate of 93.20%. The false rates of both applications are very low. The proposed indirect trust inference method is efficient, and it outperforms the existing approach.