Mathematical Foundations of Computer Security

Machine learning based anomaly detection method for SQL

A. I. Murzina

Tomsk State Uneversity, Department of Informatics, Tomsk

Abstract: In this paper, an anomaly detection method for SQL is proposed. The method is based on the clasterization and recurrent neural networks for legitimate SQL-queries. The main idea is to teach neural network to detect non-typical SQL-queries for the server including queries independent from known instances of successful attacks.

Keywords: machine learning, anomaly detection, SQL-injections, clasterization, recurrent neural network.

UDC: 004.94

DOI: 10.17223/2226308X/10/46