/Network-Anomaly-Detection-using-Autoencoders

This repository contains an implementation of an Autoencoder-based Neural Network designed to detect anomalies in network traffic, using the NSL-KDD dataset.

Primary LanguageJupyter Notebook

Network-Anomaly-Detection-using-Autoencoders

Embark on a cybersecurity journey with our PyTorch-based neural network classifier, specifically designed for NSL-KDD data. This project focuses on the careful preparation of NSL-KDD data, including normalization of numerical attributes and one-hot encoding of categorical attributes.

Key Features:

  1. Data Preparation: Streamline the preprocessing of NSL-KDD data by normalizing numerical attributes and implementing one-hot encoding for categorical attributes.

  2. Neural Network Design: Delve into the core of the project with a PyTorch-based neural network that emphasizes the construction of encoders and decoders.

  3. Hyper-Parameter Optimization: Model performance was enhanced by fine-tuning hyper-parameters, such as the learning rate and reconstruction error threshold.

  4. Performance Evaluation: The neural network attains a 89% accuracy on the NSL-KDD validation dataset.

Browse the repository to examine the code.