Cost-effective optimization of data dependency based intrusion detection system

The existing models of evaluating and optimizing the cost efficiency of intrusion detection have limited applicability in terms of the database intrusion detection methods. Database management systems have peculiarities significantly affecting the cost effectiveness of those methods. This thesis is focused on solving the problem of cost-effective optimization in systems that use dependencies among data items to detect malicious transactions. The model suggested in the thesis considers three main factors: the quality of intrusion detection, the probability of intrusion and the cost structure of a company whose data is protected by the intrusion detection system. We estimate these factors while taking into account such specific features of databases as the mandatory authentication of all users and the ability to rollback a transaction.