PASM : a Policy Aware Social Miner

[摘要] The Policy Aware Social Miner (PASM) project focuses on creating awareness of how seemingly harmless social data might reveal sensitive information about a person, which could be potentially abused. It seeks to define good practices around social data mining. PASM allows people to create policies governing the use of their personal information on social networks. Using linked data, PASM semantically enhances the usage restrictions to ensure that potentially sensitive information is identified and appropriate policies are enforced. PASM also enables people to provide refutations for other information about them that is found on the Web. PASM encourages consumers of social information on the Web to use the mined data appropriately by enforcing data policies before returning the search results. PASM provides a solution to the following issue of privacy in social data mining - although people know that searches for data about them are possible, they have no way to either control the data that is put on the Web by others or indicate how they would like to restrict use of their own data. In a user study conducted to measure the performance of PASM in identifying sensitive posts as compared to the study participants, PASM obtained an F-Measure of 84% and an accuracy of 80%. Interestingly, PASM demonstrated a higher recall than precision, a property that was valued by the study participants as all but one participant indicated that they would prefer receiving false positives rather than false negatives.

[发布机构] Massachusetts Institute of Technology