KMFC-GWO: A Hybrid Fuzzy-Metaheuristic Algorithm for Privacy Preserving in Graph-based Social Networks

Abstract

In recent years, the proliferation of social networks has been remarkable, providing a rich source for data mining endeavours. However, a significant challenge lies in safeguarding the privacy of individuals while sharing these databases publicly. Current approaches such as K-anonymity, L-diversity, and Tcloseness, are commonly employed for data anonymization in social networks. However, these techniques entail considerable information loss due to random alterations in the graph-based datasets. To addressthese limitations, this paper introduces a new anonymization technique called KMFC-GWO, which combines K-Member Fuzzy Clustering with Grey Wolf Optimizer. This integrated method is designed to strengthen the anonymized graph against a range of threats, including identity, attribute, link disclosure, and similarity attacks, while significantly reducing information loss. Within the KMFC-GWO framework, Kmember fuzzy c-means clustering is utilized to create wellbalanced clusters, each meeting the K-anonymity requirement. Subsequently, the Grey Wolf Optimizer is applied to optimize cluster formation and effectively anonymize the social network graph. The objective function is carefully crafted to minimize both clustering error and information loss, while ensuring adherence to predefined anonymity criteria.