Machine Learning Based Network Censorship
Xiangyu Gao, Meikang Qiu, Meiqin Liu · 2021
Network censorship is to actively modify network users' experience in order to restrict behaviors to some extent. One of the most popular network censorship actions is to add several IP addresses or domain names into the blacklist and prevent users from visiting those websites. However, this blacklist should not be static since most web developers, after finding their websites are in the blacklist, would like to transplant their content to other IP addresses or domain names and redirect network users to new addresses. In response to these concerns, in this paper, we propose a machine learning based network censorship design which can update the blacklist by analyzing network users' behavior. By leveraging data analysis for the network users' behaviors when facing network censorship, we collect a lot of useful information and utilize it to add or delete members from the blacklist. We think this network censorship design can be more effective than static censorship with respect to filtering forbidden information and retaining qualified candidates.