Privacy Preserving Schemes for Secure Interactions in Online Social Networks

Online Social networks (OSNs) play a major part in everyone's daily life. The Sybil uses the original identity of targeted or random victim to create an account on the social network. All these effects should be removed only by mitigating the creation of Sybil accounts. The existing techniques for Sybil detection employs reactive defence strategies which are initiated based on the user report, where by the time, the user becomes victim. In this paper, a novel Sybil detection framework is proposed to identify the attempt of account creation by the Sybil users during the admission phase and also making the defending pattern unpredictable to learn. The Sybil detection framework consists of a three level privacy scheme, which is capable of strongly suspecting the fake from the genuine. The extensively used open-source, distributed computing platform, Apache Hadoop with its ecosystem tools are utilized to implement the Sybil detection framework for effective processing of unstructured social data. The experiment is conducted with data collected from Twitter, Facebook and Google+. Experimental results show that proposed Sybil detection framework is well suited for mitigating the worst effects of fake users with efficient handling of large social data. The K-means clustering and Bayesian classification are used to detect the Sybil accounts with 94% accuracy and 92% precision using many public features.