DIPP: Information Diffusion for Privacy in Multi-agent Systems

Abstract

Ensuring the privacy of users is a key component in collaborative computer systems, where users can access private information of others. Social networks are a prime example of such systems. Therefore, privacy has to be ensured not only by the administrators but also by users in collaboration. The content users choose to share may conflict with the privacy preferences of their own or those of others, given the context of the content. Thus, a decision to share or not to share can be seen as a privacy decision. To manage privacy preferences better, it is important to understand how they appear and disappear on social networks. However, it is also important to understand how the privacy preferences spread throughout the network. Given this understanding, one can reason about the implications of the spreading mechanism has on mitigating or promoting certain privacy preferences.

In this work, the diffusion of infectious privacy preferences (DIPP) model is proposed to investigate how these privacy preferences spread on social networks. An epidemic model is used to model the spread of privacy preferences. Simulations of social network interactions are used to investigate various circumstances, such as the rarity of a privacy preference and opposition of a privacy preference, and their effect on the diffusion of privacy preferences. Furthermore, we investigate the effect of using a trust model to model trust between agents. The results show that the DIPP provides a stable foundation to further research the spread of privacy preferences in online social networks.