Applying accuracy-based LCS to detecting anomalous database access

Suin Seo, Sung-Bae Cho

Abstract: Database intrusion detection (DB-IDS) is the problem of detecting anomalous queries in transaction systems like e-commerce platform. The adaptive detection algorithm is necessary to find anomaly accesses when the environment changes continuously. To solve this problem, we used accuracy-based LCS (XCS), one of the primary model of adaptive machine learning method, for detecting malicious accesses in databases. In the problem of database intrusion detection which changes the detecting targets, we found and analyzed the patterns of rule generation to show systemically how the adaptive learning of XCS algorithm is working in practical usage.