Data represent today an important asset for companies and organizations and must be protected. Most of an organization's sensitive and proprietary data resides in a Database Management System (DBMS). Various attacks (e.g., malicious transactions) may corrupt data items in the database systems, which decreases the integrity level of the database. Malicious transactions detection mechanisms are becoming more and more sophisticated to detect such attacks. These mechanisms are implemented either externally as an autonomous subsystems separated from the DBMS, internally to the DBMS using database triggers, or internally to the DBMS using database stored procedures by compiling them into native code residing in shared libraries. The focus of this paper is to investigate the effect of malicious transactions
detection mechanisms implementation on database performance. This paper presents implementation of three mechanisms for detection of malicious transactions in the Oracle 10g DBMS, and investigates the performance of the three mechanisms using a telephone database. The experimental results showed that the average performance overhead caused by the activation of the external and the native mechanisms is about 40%, and about 48% for the database trigger mechanism. As a result, the external and the native mechanisms outperform the database trigger mechanism in term of database performance.