Optimized Detection of SQL Injection Attacks with Layered Decision Forest

The attacks related to SQL injection are on the top list of security threats, and their complexity with rapid evolution makes it damaging the database, resulting in data breaches and website downtime. Because of the wide range of attack payloads, diversity of strategies, and varied execution modes, detecting SQL attacks is still challenging. This proposed work uses a Layered Decision Forest (LDF) with AdaBoost to address these problems. The LDF framework optimizes feature processing as the raw feature vector and averaged outputs from earlier layers are fed into each successive layer to minimize the degradation of features in deeper models. In addition, an AdaBoost-enhanced LDF was proposed to adjust feature weights dynamically during training according to error rates so that the model can better process multidimensional features and avoid overfitting. The results show that LDF is more accurate and robust than the traditional ML and DL techniques for SQL injection detection.