search

Lovish-Dak / Network-Intrusion-Detection-System

In this project, I created a network intrusion detection system using CNN and BiLSTM layers. I trained the model on NSL-KDD & UNSW-NB15 dataset. I used some pre-processing techniques such as Min-Max Normalization, One-Hot Encoding etc. and measured the performance of the model on metrics such as accuracy, FPR, FNR, Precision, Recall & F-1 Score
6 stars 0 forks source link

readme

Network-Intrusion-Detection-System

Overview

This is my B.Tech Project, where I created a model which can detect network intrusions effectively. For the effective intrusion detection system, I used the Convolutional Neural Networks as they can capture both spatial and temporal patterns, and I used BiLSTM layers as well because LSTM generalizes well on time series or sequential data.

Datasets Used

(i) NSL-KDD
(ii) UNSW-NB15

Techniques Used

(i) Removing NULL and duplicate values
(ii) Oversampling
(iii) One-Hot Encoding
(iv) Min-Max Normalization
(v) Stratified K-Fold Cross Validation

Model Architecture (Layers Used)

  1. Conv-1D Layer
  2. Max Pooling Layer
  3. Batch Normalization Layer
  4. BiLSTM Layer
  5. Reshape Layer
  6. Max Pooling Layer
  7. Batch Normalization Layer
  8. BiLSTM Layer
  9. Dropout Layer
  10. Dense Layer (Softmax Regression in case of Multi-Class Classification and Logistic Regression in case of Binary Classification)

PPT Link: https://www.canva.com/design/DAGNOmyvUU4/5tyNhD6c0KHF3Fx624fEnQ/view?utm_content=DAGNOmyvUU4&utm_campaign=designshare&utm_medium=link&utm_source=editor