Inference attacks based on neural networks in social networks

Abstract

In modern society, social networks play an important role for online users. However, one unignorable problem behind the booming of the services is privacy issues. At the same time, neural networks have been swiftly developed in recent years, and are proved to be very effective in inference attack. This paper conducts an extensive study to infer sensitive personal information from public insensitive attributes in social networks by deploying fully connected neural networks. Correlation matrices and the details of constructing neural networks for social networks are elaborated. To show the advantages of neural networks on inference attack, different traditional machine learning algorithms are also studied. The results show that neural networks can achieve about 4 times of the baseline accuracy to classify low-correlation, high-noise dataset to infer sensitive users' attributes. In addition, neural networks outperform all the selected traditional algorithms. Outcomes from the study are deliberately discussed, and the limitations of both neural networks and traditional machine learning algorithms are also illustrated.