Privacy Analysis And Protection In Mobile Social Network

Nowadays,social networks are overwhelmingly popular and various services attract billions of users.In order to provide services such as making friends,personalization,etc.,social networks collect large amount of user information and behaviors.However,the privacy issues caused by these information and behaviors also bring many concerns and become a hot research topic.The paper aims at studying privacy issues from social networks and proposing corresponding solutions.Along with the popularity of mobile social networks,privacy leakage risks caused by location sharing are becoming non-negligible.This paper firstly compare the shared locations on mobile social networks with ground truth traces to quantitatively evaluate privacy leakage due to location sharing.In order to achieve this goal,30 volunteers are hired to conduct a 3-week experiment.Furthermore,this paper discovers whether more privacy can be inferred given shared locations.Our experiments show that users' demographics(e.g.,gender,education)can be accurately estimated given shared locations.To address these issues,this paper proposes a context aware system level location privacy protection framework aiming at learning users' location privacy preferences and providing a transparent privacy control system.On the other hand,though widely used for personalized recommendation and advertisement delivery,user profile aggregation from multiple social networks inevitably cause serious privacy risks.From privacy's perspective,de-anonymization attacks are well studied.This paper combines merits of previous approaches and proposes a novel and practical de-anonymization scheme,which reflects severity of the problem.The proposed scheme first leverages social network graph to find paired communities so that the candidate set can be reduced,then use public available profile information to map accounts with a high precision.This approach is evaluated using real-world datasets and the privacy leakage is quantified.Results show that much information will be leaked after de-anonymization and more solutions for this problem are needed.