SybilSort algorithm - a friend request decision tracking recommender system in online social networks

摘要

Sybils are detected and prevented from making friends on Online Social Networks (OSN) by the proposed SybilSort algorithm. SybilSort operates in the following manner: Task 1 (OSN registration): In order to register with OSN, a person must upload a Government Identity (GI). Task 2 (sort model): OSN is randomly clustered into equal-sized clusters. When one person receives a friend request from another, the decision to accept or reject is based on the dynamic thresholds determined by the cluster members and Cluster Monitor (CM). The person can then decide whether to accept or reject the request. If the variation is detected, the CM tracks the decision taken despite the recommendation, and the second suggestion is presented as an alert, along with proof of convergence, Sybil similarity, and priority assignment scores. Task 3 (cluster shuffle): The cluster is examined, and if there are any Sybil activities, the cluster is shuffled. To establish cluster shuffle, the cumulative difference is calculated using Belief Level Deviation (BLD), the Jaccard coefficient, and the Hamming distance. The research goal is to avoid the creation of the Sybil profile, to remove vote collisions, to track decisions after recommendation, and to avoid Sybil density. The proposed SybilSort is compared to existing Sybil detection algorithms using performance evaluations for varying numbers of persons and average network sizes. Using the 2019 Facebook dataset, the precision, recall, F1-score, and Region Of Curve (ROC) metrics are examined. The final results reveal that the proposed SybilSort algorithm significantly outperforms the already validated Sybil detection algorithms.