Preserving privacy of users in online social networks is important. Usually, users specify their privacy constraints and the online social network is expected to enforce them. However, many times a piece of content is related to a number of users, whose privacy constraints might be incompatible. For example, when a user shares a party photo, her privacy constraints are enforced on the picture but the privacy constraints of the people in the picture are not. One way to deal with this problem is to enable collaborative policies to be written per content [5]. However, composing privacy policies from scratch is extremely time consuming. Further, it is difficult to overcome conflicts among users. Another way of dealing with this problem is to use agent-based approaches such that agents represent their users and employ agreement techniques to reach a conclusion on whether a content should be shared or not [3, 4]. Following this line of work, we advocate an agent-based approach where each user in the social network is represented by an agent that manages its user's privacy constraints. Each agent represents its domain knowledge using an ontology and its user's privacy constraints as semantic rules. When a user wants to post a content, her agent contacts the relevant agents (e.g., agents of tagged users in the content) to request permission. Upon receiving the request, these agents evaluate it using their own rules. If any of the agents has a concern; i.e., its privacy constraint is violated, then the agents engage in an argumentation session. The argumentation is done in a distributed manner where agents take turns to provide evidence as to why the content should be shared or vice verse. The evidence is generated from the agent's ontology and semantic rules on demand based on what the other agents have proposed. At the end of the argumentation, winning arguments are computed, leading to a decision on whether the content should be shared or not.
展开▼