Intelligent WAF
Guarding Beyond Boundaries: Intelligent WAF - Your Shield! Rate Limiting, Port Filtering, IP Blacklisting, and Next-Gen Deep Learning Defense Against the OWASP Top 10 Menaces.
Created on 28th January 2024
•
Intelligent WAF
Guarding Beyond Boundaries: Intelligent WAF - Your Shield! Rate Limiting, Port Filtering, IP Blacklisting, and Next-Gen Deep Learning Defense Against the OWASP Top 10 Menaces.
The problem Intelligent WAF solves
The Web Application Firewall (WAF) is designed to protect web applications by monitoring, filtering, and blocking HTTP traffic between a web application and the Internet. It exists in between the client and the server endpoint. It is a deep learning based Web Application Firewall by leveraging deep learning frameworks. This WAFs can detect and thwart zero-day attacks, minimizing false positives, and adapting to evolving threats more effectively. They excel in identifying complex attack patterns and offer deeper insights into web traffic behavior. Our Deep learning based WAF streamline the security process by automating rule updates and reducing manual tuning requirements. Ultimately, they provide a more intelligent and adaptive defense mechanism for web applications, enhancing overall security posture and resilience against modern cyber threats.
Challenges we ran into
We have ran into many real world problems while developing the WAF(Web Application Firewall). Especially in the machine learning scope of the project. Some of our challenges include,
Data Collection: This firewall project requires a lot of input to train the model. Gathering the data and processing it to fit into our model was really a challenging task. We used various online platforms for gathering the datasets to train the model such as kaggle, github, deep web.
Model integration: Integrating the Machine learning model with the actual firewall was a big task since we have to develop a whole architecture and fit our model into it. The whole idea of integrating the model between server and the firewall endpoint was really challenging.
Model compatibility: We developed two ML models for the better filtering of traffic . The first model detects if the traffic is malicious or not. The Second model classifies the traffic depending on the false positive feedback from the first model
Tracks Applied (1)
Auth0
Auth0