IntrusiGen

TEAM: Undergraduate Students from COMSATS University Islamabad, Abbottabad Campus - Pakistan

Syed Shah Hussain B.	Mahad Wajid	Abdul Aziz	Daim Ahmed

Inspiration

The idea for IntrusiGen was born from the frustration of seeing countless organizations and individuals fall victim to cyber attacks simply because their security systems weren’t equipped to handle evolving threats. Traditional intrusion detection systems often struggle with skewed datasets, making it difficult to detect anomalies accurately. This imbalance leaves networks exposed to sophisticated attacks, causing devastating breaches, financial losses, and compromised trust.

We wanted to create something that could level the playing field—a solution that not only addresses data imbalance but also offers real-time, actionable insights. By combining deep learning with advanced data generation and visualization techniques, we aimed to build a system that empowers users to proactively detect threats before they can cause harm. IntrusiGen was crafted with the vision of making intelligent, automated security accessible to everyone, regardless of resources or expertise.

What it does

IntrusiGen is an AI-powered Intrusion Detection System designed to simplify and enhance network security through an intelligent, automated process. It offers users a seamless experience from dataset upload to real-time threat detection and visualization. Here's what it does:

1️⃣ Data Upload & Visualization:

Users can effortlessly upload network traffic datasets. IntrusiGen immediately visualizes critical insights, including the number of classes, feature distribution, and overall dataset composition, making it easy to understand the dataset's structure.

2️⃣ Data Preprocessing & Balancing:

To ensure high detection accuracy, the system applies advanced preprocessing techniques to clean and structure raw data. It then leverages Generative Adversarial Networks (GANs) to generate synthetic samples, effectively balancing skewed datasets and enhancing robustness.

3️⃣ Real-Time Threat Detection:

The balanced dataset undergoes rigorous analysis using an XGBoost classifier, which identifies anomalies and cyber threats with high precision. This real-time detection mechanism ensures prompt recognition of malicious activities.

4️⃣ Actionable Insights & Reports:

IntrusiGen provides users with clear, actionable insights through data visualizations and detailed reports, empowering them to strengthen their network defenses effectively.

5️⃣ User-Friendly Interface:

Designed with simplicity in mind, the platform is accessible to beginners and experts alike, making advanced cybersecurity tools available to all.

Dataset we used - UNSW-NB15

How we built it

Building IntrusiGen was a collaborative effort, combining expertise in AI, machine learning, web development, and cybersecurity.

1. Research & Planning:

We began by identifying common pain points in existing Intrusion Detection Systems (IDS), particularly their struggles with imbalanced datasets and inefficient anomaly detection. We then designed a solution incorporating cutting-edge AI technologies to address these issues effectively.

2. Frontend Development (React.js):

We built a clean, user-friendly interface using React.js, ensuring seamless dataset uploads and interactive visualizations. Chart.js, React-chart.js-2, and Victory were used to generate insightful graphs and charts that help users easily understand dataset structures and results.

3. Backend Development (Node.js & Python Shell):

Our backend architecture, powered by Node.js and Python Shell, handles data preprocessing, model training, and real-time predictions. The integration of Python Shell allows us to leverage powerful machine learning libraries while keeping the backend structure efficient.

4. Data Processing & Generation (GANs):

To tackle data imbalance, we employed Generative Adversarial Networks (GANs) to generate synthetic samples of minority classes. This approach significantly improved the robustness of our detection system. Essential Python libraries like Pandas, numpy, torch, and Sklearn were used for data processing and model implementation.

5. Threat Detection & Classification (XGBoost):

We trained an XGBoost classifier on the balanced dataset, fine-tuning its hyperparameters to achieve high accuracy in detecting cyber threats. The model's real-time analysis capability ensures immediate identification of malicious activities.

6. Database Management (MongoDB):

Processed datasets and user-generated reports are securely stored in a MongoDB database, providing easy retrieval and storage.

7. Deployment & User Accessibility:

IntrusiGen was deployed as a web application, accessible to users of all levels—whether they are cybersecurity experts or beginners looking to enhance their network protection.

Architecture

Architecture Picture

Challenges we ran into

• Handling heavily skewed datasets was a major challenge. Without proper balancing, our model's accuracy was severely impacted.
• Ensuring real-time detection and visualization without compromising accuracy or speed was a complex task.
• Developing clear, informative visualizations for dataset insights and detection results.

Accomplishments that we're proud of

• Successfully balanced skewed datasets using GANs for enhanced detection accuracy.
• Developed a real-time detection system for instant threat identification.
• Integrated Node.js and Python Shell seamlessly for backend efficiency.
• Addressed real-world cybersecurity challenges effectively.

What's next for IntrusiGen

🌐 Real-World Implementation:

Expanding IntrusiGen beyond datasets to real-world applications, including IoT devices, routers, and enterprise networks.

📡 Real-Time Insights & Alerts:

Integrating real-time monitoring systems to provide instant alerts and actionable insights during intrusion attempts.

🔧 User-Centric Features:

Adding customizable dashboards, detailed reports, and APIs for seamless integration with existing security frameworks.

📱 Mobile & Cloud Integration:

Making IntrusiGen accessible through mobile apps and cloud platforms for broader accessibility.

Built With

• chart.js
• express.js
• mongodb
• node.js
• numpy
• pandas
• python-shell
• react.js
• scikit-learn
• torch
• victory