Privacy Wizards for Social Networking Sites Reporter :鄭志欣 Advisor: Hsing-Kuo Pao Date : 2011/01/17 1 2 Reference Lujun Fang and Kristen LeFevre. "Privacy Wizards for Social Networking Sites." 19th International World Wide Web Conference (WWW2010,Best student paper). Lujun Fang, Heedo Kim, Kristen LeFevre, Aaron Tami ,"A Privacy Recommendation Wizard for Users of Social Networking Sites" 17th ACM conference on Computer and communications security (ACM CCS2010,Demo). www.eecs.umich.edu/dm10/slides/fang.pptx 3 Outline Introduction Wizard Overview Active Learning Wizard Evaluation Conclusion 4 Introduction Social network sites have been increasingly gaining popularity. More than 500 million members Privacy is a huge problem for users of social networking sites. More Personal information A lot of Friends (Ex: FB average 130) Facebook’s “Privacy Setting” is too detail. 5 Goal We propose the first privacy wizard for social networking sites. The goal of the wizard is to automatically configure a user's privacy settings with effort from the user. 6 Challenges Low Effort , High Accuracy Graceful Degradation Visible Data Incrementality 7 Idea Idea: With limited information, build a model to predict user’s preferences, auto-configure settings 8 Wizard Overview 9 Active Learning Wizard Classifier Each friend as a feature vector Question How to extract features from friends? How to solicit user input? 10 Extracting Features Age Sex G G G G G G 0 1 2 20 21 22 G3 Obama Pref. Label Fan (DOB) (Alice) 25 F 0 1 0 0 0 0 0 1 allow (Bob) 18 M 0 0 1 1 0 0 0 0 deny (Carol) 30 F 1 0 0 0 0 0 0 0 ? G0 G1 G21 G2 G3 G20 G22 11 Soliciting User Input Ask Simple and Right questions Question : Would you like to share your Date of Birth with ? How to choose informative friends using an active learning approach? Uncertainty sampling 12 Figure 5: Screenshot of user study application , general questions Figure 6: Screenshot of user study application,detailed questions. 13 14 Evaluation Gathered raw preference data from 45 real Facebook users. How effective is the active learning wizard, compared to alternative tools? 15 Experiments DTree-Active Model is a Decision tree Uncertainty sampling Decision Tree Model is a Decision tree User labels randomly selected examples Brute-Force Like Facebook policy-specification tool Assign friends to lists Result 16 17 Tradeoff 18 Conclusion Privacy is an important emerging problem in online social networks. This paper presented a template for the design of a privacy wizard, which removes much of the burden from individual users.