Feature Ratio Method: A Payload Feature Extraction and Detection Approach for SQL Injection Attacks

Due to the complexity of the deployment environment and the differences in the syntax of the backend database engine, SQL injection attacks have the characteristics of high complexity and poor timeliness. In recent years, the ability to detect and predict SQL injection attacks based on machine learning has been widely recognized in preventing unknown network security risks. However, most of these studies use the method in text classification to extract features of SQL injection attack payloads. The SQL injection attack payload is different from ordinary natural language, so there are problems with inaccurate segmentation and large vocabulary in the conventional feature extraction methods. This paper proposes a feature ratio method for feature extraction and detection of SQL injection attack payloads and compares it with various machine learning algorithms. The experimental results show that the average F1 Score of this method reaches 96.29%, which is better than the methods of the previous studies in terms of accuracy and precision. It shows that our method can better help machine learning algorithms to classify and detect SQL injection attack payloads.