Emergent Privacy Norms for Collaborative Systems

Managing privacy of users in online systems is a major aspect of cyber-security. Typical approaches to privacy are concerned with giving users options of informed consent, wherein users define their private data, how they want them to be used, and so on. However, in collaborative systems, such as online social networks, managing privacy exhibits problems beyond traditional consent, since a content being shared (such as a group picture or a multi-party business contract) might belong to more than a single entity, with different privacy policies. Recent approaches to preserve privacy in such settings rely on multiagent agreement technologies, which require a new decision to be formed for every content that will be shared, making them difficult to scale for real life applications. Accordingly, this paper proposes a normative approach for maintaining privacy in collaborative systems that do not require a decision to be formulated from scratch for each content. Instead, the system generates social norms based on previous decisions. The agents are free to follow the social norms as well as their own privacy policies. We show over multiagent simulations that our approach extracts social norms successfully and enables successful privacy decisions to be taken.