Association rule based data mining approaches for Web Cache Maintenance and adaptive Intrusion Detection systems

The efficiency of real time decision making systems, configured with a set of rules for decision making strongly depends on the stability of the rule sets maintained. Hence, due to changing external and internal factors, adaptive systems which can redefine rule sets to maintain optimal performance are needed. We have used an Association Rule based data mining approach to dynamically identify rule changes and quickly update the rule set to maintain optimal performance consistently on two real-time problem areas namely Web Cache Maintenance and Intrusion Detection. In the first problem, a multi-agent approach performs optimized rule set formulation; performance calculation and rule set adjustments for maintaining an optimal performance of proxy Web caches. In the second problem, an adaptive data dependency rule set maintenance approach was adopted to give a foolproof malicious transaction detection system for web accessed large databases based on access role violations and database cache penetration.