| dc.description.abstract | Pornography on the web brings non-ignorable influence on minors. Moreover, according to statistics, the problem is hard to solve by simply asking parents and teachers to watch their children go on line. Web filtering, on the other hand, is a potential choice to mitigate the problem. The web filter for pornography relies on expensive application layer analysis, and the rise of Cloud Computing provides a direction for proposing a Cloud-based web filtering architecture to cut the cost. Nevertheless, the architecture of web filtering on Cloud causes significant delivery overhead. Besides, the compatibility problems, different deployed scenarios and requirements of users are also needed to consider for achieving web filtering on Cloud. The web filtering strategies for public Cloud scenarios are proposed in the thesis. The strategies maintain TCP session on Cloud to reduce the complexity of network equipment. It uses the “Man-in-the-middle” way instead of TCP hijacking, thus reduces the delivery overhead on Cloud, and is compatible to current web content providing services, such as transparent proxy and content delivery network. Moreover, system design of Cloud-based web filtering service is then proposed based on the proposed strategies. The system offers a flexible range of deployment options, and can provide different service level for costumers. Analysis shows that the proposed web filtering strategies can reduce delivery overhead in public Cloud scenarios, and the improve rate is inversely proportional to the blocking rate. Experiments show that the proposed system is compatible to the current network environment.
| en_US |