SaaS-oriented Context-aware Data Filtering Model And Algorithm

Software as a Service(SaaS), relying on the cloud computing, could provide service for multiple tenants via Internet at a lower price, but quicker deployment and higher configurability, which makes SaaS a representative for future's software.Since multiple tenants share access to SaaS via Internet, there's a higher security threat to SaaS than traditional one. Based on the shared database model in SaaS, the thesis gives a research on data security and data filtering, including a context-aware data filtering model on the basis of dynamic strategies, and quick and effective context matching algorithms.At first, the thesis proposes a SaaS-oriented context-aware data filtering model, which consists of two phases: strategy making and strategy execution. The model provides support for dynamic and high configurable data filtering. Following the formal definitions of context and data property, the model provides the concepts of context decision, filtering rule and filtering strategy. Context decision is used to match context value for searching corresponding filtering rules; filtering rule conducts how data filtering is executed; filtering strategy builds mapping between context decisions and filtering rules.Based on the data filtering model, two matching algorithms are proposed for the context matching. Traversing algorithm builds strategy tree with non-leaf node as context atomic decision, and traverses the whole tree in pre-order to match context. Sequential algorithm presents strategies with nodes list, sorts the atomic decisions according to their matching rate, and uses equivalent link to eliminate redundant context decisions. The sequential algorithm could greatly improve the efficiency of matching process.Next, an analysis of the key process for data filtering is presented. Filtering rules are classified into add rules and delete rules, and the potential conflict between these two kinds of rules is discussed. Finally, the thesis proposes a way to detect potential conflict using data property and avoid it with rule weight. At last, we develop a data filtering system based on the mentioned model. In the system, strategies are stored in XML format, and filtering rules are stored as SQL. The system could be used to provide dynamic fine-grained data filtering.