Detecting command injection attacks in web applications based on novel deep learning methods.

Area of Science:

• Cybersecurity
• Artificial Intelligence
• Deep Learning

Background:

• Web command injection attacks threaten web applications, causing data breaches and service disruptions.
• Existing detection methods face challenges with complex, obfuscated attacks, leading to low efficiency and poor accuracy.
• Advanced feature extraction and temporal analysis are crucial for effective detection.

Purpose of the Study:

• To propose a novel deep learning model for enhanced identification of web command injection attacks.
• To improve the accuracy and efficiency of detecting sophisticated malicious code in web applications.
• To address the limitations of traditional detection methods in handling complex attack patterns.

Main Methods:

• Developed the Convolutional Channel-BiLSTM Attention (CCBA) model, integrating deep learning techniques.
• Utilized dual CNN convolutional channels for comprehensive feature extraction.
• Employed a BiLSTM network for bidirectional temporal feature recognition and an attention mechanism for feature weighting.

Main Results:

• The CCBA model achieved 99.3% accuracy and 98.2% recall on a real-world dataset.
• Consistent performance with over 98% accuracy on two public cybersecurity datasets, demonstrating robustness and generalization.
• Outperformed existing methods in identifying web command injection attacks.

Conclusions:

• The CCBA model presents a highly effective solution for detecting web command injection attacks.
• Deep learning, particularly the CCBA architecture, significantly enhances web application security.
• The model's superior performance validates its potential for real-world cybersecurity applications.