Research on SQL Injection Detection Based on Deep Learning

With the rapid development of modern information technology, web applications have gradually become important service providers in people's daily work and life. However, while web applications bring convenience, they also face various challenges. Among them, SQL injection attacks, characterized by their diversity, rapid changes and strong concealment, have always been one of the main threats to web applications. Currently, most mainstream SQL injection detection tools on the market are based on established rules and cannot cope with the constantly changing challenges. Therefore, this paper proposes a deep learning-based approach using pretraining and fine-tuning with BERT for SQL statement pretraining. An improved TextCNN algorithm is used as the classification algorithm. Experimental results show that the proposed algorithm has significant improvements compared to other text vectorization algorithms in terms of detection performance and has certain application prospects.