CyberGuard

🔗 Deployed Link: CyberGuard

Tech Stack

• Python
• Gradio
• PyTorch
• Scikit-Learn
• Transformers
• Accelerate
• NLTK
• NumPy
• Pandas

Features

Data Handling

• Cybercrime complaints are preprocessed using Pandas.
• Custom PyTorch Dataset class used for loading data.

Model Architecture

• Fine-tuned BERT (BertForSequenceClassification) model for multi-class text classification.
• BertTokenizerFast used for tokenizing inputs efficiently.

Inference Pipeline

• Hugging Face pipeline used for quick and clean inference setup.
• Runs seamlessly on CPU or GPU with Torch backend.

Interface

• Deployed via Gradio for an interactive user experience.

Workflow

1. Data Preparation
   Load CSV data with Pandas, clean and encode labels.
   Define custom dataset class using torch.utils.data.Dataset.

2. Model Training
   Fine-tune BertForSequenceClassification with PyTorch.

3. Inference
   Use Hugging Face pipeline for streamlined inference deployment.

4. Deployment
   Gradio app allows users to input a complaint and see classification results instantly.