Database security is a crucial concern today. One mechanism for safeguarding information stored in database systems is to use an Intrusion Detection System (IDS). Recently researchers are working on using machine learning techniques to increase the accuracy of the detection malicious attacks on database systems; Such as mining data dependencies among data items, access patterns of users and learning SQL commands. In this paper, the authors survey some intrusion detection approaches, which use these techniques. Also, they discuss the advantages and disadvantages of the approaches and compare them with considering their different features.