Deep Learning Technique-Enabled Web Application Firewall for the Detection of Web Attacks.

Area of Science:

• Computer Science
• Cybersecurity
• Network Security

Background:

• Web applications face increasing threats from unauthorized access, data theft, and destruction.
• Traditional network firewalls struggle to protect against sophisticated attacks like Distributed Denial of Service (DDoS), SQL injection, and Cross-Site Scripting (XSS).
• Effective detection and mitigation of diverse cyber threats require advanced analytical approaches.

Purpose of the Study:

• To conduct a comparative analysis of normal HTTP traffic versus attack traffic to identify key indicators.
• To develop and evaluate a layered architecture model for detecting DDoS, XSS, and SQL injection attacks.
• To enhance web application security through intelligent traffic analysis and filtering.

Main Methods:

• Comparative analysis of normal and attack HTTP traffic using standard datasets (ISCX, CISC, CICDDoS).
• Development of a Long Short-Term Memory (LSTM)-based layered architecture for attack detection.
• Implementation of a two-layer model: DDoS detection followed by XSS and SQL injection detection.
• Integration with a Web Application Firewall (WAF) for application-level security.

Main Results:

• The first layer (DDoS detection) achieved an accuracy of 97.57%.
• The second layer (XSS and SQL injection detection) achieved an accuracy of 89.34%.
• The model effectively filters high-rate HTTP traffic before deeper analysis.

Conclusions:

• The proposed layered architecture demonstrates significant effectiveness in detecting multiple web attack vectors.
• LSTM networks provide a robust foundation for building accurate cyber threat detection systems.
• The integration of this model with WAF enhances overall web application security against sophisticated attacks.