Privacy Leakage and Quantitative Analysis in Social Network Big Data

With the development of social networks, multiple user data not only creates opportunities for data-driven applications, but also brings a serious privacy breach risk. Research now always focuses on qualitative description and lacks a systematic quantitative assessment framework. This article uses some open data sets like Twitter and Facebook, which model three typical attack scenarios--attribute reference, linkage predictions, and model reversals--and rates the anonymous, differential privacy, mixed defense, and federal study. The result of experiments says that, although in an anonymous situation, privacy hazards are still notable. Attackers can refer to sensitive information from the structure and behavior features. Although anonymity decreases the attack mission rate, distinct utility losses always occur. Mixed defense realizes a better balance. Federal studies have the ability to resist model reversal and face the challenge of expansion. Further analysis shows that cross-platform datasets and the activity of users magnify the privacy risk. In conclusion, this research not only creates new strategies in qualitative privacy leaks but also has a balance in data effectiveness and privacy protection between social platforms and regulators.