Research On Methods Of Damaged Data Detection Based On Application Semantics

The database management system is the most central part of the information system,most important information were stored in the database, however, it often incurs hackerattacks and thus results significant losses to users. Intrusion tolerant database technologyis designed to respond to the intrusion has occurred, to ensure that the database system tocontinue to provide services from malicious attack. As a key sub-module of the invasiontolerate database, the damaged data detection were used to detect the infectedtransactions after the attack happened.By analyzing the deficiencies of traditional damaged data detection method,application semantics were introduced into the traditional method to improve the resultsof the damage assessment, and on this basis, the architecture of the damaged datadetection system based on application semantics was designed. The system is dividedinto the traditional transaction dependency module, application semantics module anddamaged estimation module. Traditional transaction dependency module accordancetransaction dependent model to determine and record dependencies between transaction.The application semantics module uses semantic technology to handle the transactiondependencies from traditional transaction dependency module. The damaged estimationmodule analyzes all infected transactions by attack transaction.Introduce the key technology in the damaged data detection system based onapplication semantics. Transaction dependency graph-based tracking technology to trackthe dependencies between transactions, and to increase system table save transactionsdependencies. To complete the definition of the application semantics in the context ofdatabase systems, then using semantic recognition technology to convert the low-leveldependency between transactions to the senior dependency between semantics, andcutting invalid transaction dependency by the advantage of the increased businessrelationship table. Using damaged estimation algorithm to calculate all infected transaction by attack transaction.The experimental results show that, the method of damaged data detection based onthe application semantics is superior to the method of traditional damaged datadetection. In terms of accuracy, the method based on semantics could not only ensurethe completeness of the methods of traditional detection, but also could dramaticallyreduce the false positive error in traditional method; besides, with regard toperformance, the method based on semantics reduces the space overhead and slightlyincreases the time overhead. Finally, the experimental results were analyzed.