Statebert: Enhancing Bert with State Machine for SQL Injection Detection

Aminu Tukur1, Kabir Umar2 and Aliyu Isah Agaie3

1Department of Computer Science, Faculty of Computing, Bayero University Kano
2Department of Software Engineering, Faculty of Computing, Bayero University Kano
3Department of Information and Media Studies, Faculty of Communication, Bayero University Kano

*Corresponding author’s Email: Tukuraminu85@gmail.com, doi.org/10.55639/607.090807


ABSTRACT

SQL injection vulnerabilities continue to be a significant threat to web
applications as shown in OWASP 2021 Ranking. It enables attackers to gainunauthorized access to sensitive data and potentially execute malicious code on the server. Traditional signature-based detection methods often fail to identify novel or obfuscated SQL injection attacks. This paper presents a StateBERT approach that combines the strengths of a state machine and a BERT model to enhance the detection and mitigation of SQL injection vulnerabilities. The state machine handles the structural analysis of SQL queries, while the BERT model provides advanced language understanding to identify more complex, context-dependent anomalies. The paper discusses how this enhanced approach can effectively handle various types of SQL injection vulnerabilities. By leveraging the complementary capabilities of the state machine and the BERT model. The paper also presents the Proposed experiment, expected results and future directions. The Preliminary result is promising.

Keywords:

SQL injection,
State Machine,
BERT,
Vulnerabilities