Abstract

With the coming up of social network sites like Twitter, Facebook, LinkedIn, etc information sharing has seen tremendous growth in the past few years. Although people wish to interact with like-minded persons through such platforms, there is a lot of personal information about individuals that need to be secured. To maintain the privacy of sensitive information in a social network, efficient privacy-preserving techniques like k-anonymization, randomization, generalization, etc., are used. The majority of the proposed techniques apply edge editing and clustering based methods for creating k-anonymized social network graphs. Since these techniques distort the original properties of the graph to a great extent, various enhancements have been proposed which consider the addition of noisy nodes as one of the alternatives. In this paper, an improved version of k-degree-anonymization on social network graph has been provided which uses the hybridization of Neural Network and SVM, called NeuroSVM where average path length of the graph is preserved and addition of noisy nodes and noisy edges is reduced significantly. The proposed technique is evaluated on various parameters which include Average path length, Precision, Recall, F-measure, and Information loss. It has been experimentally proved that the proposed technique has less distortion in average path length as compared to existing techniques. The accuracy of the proposed technique is more than 75% and information loss is also reduced.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call