Skip to content

README.md

Latest commit

 

History

History
73 lines (50 loc) · 3.05 KB

README.md

File metadata and controls

73 lines (50 loc) · 3.05 KB

Deep Neural Network Implementation for Ransomware Family Classification

Author: Leo Martinez III - LinkedIn

Contact: leo.martinez@students.tamuk.edu

Created: Summer 2024

To clone this repository:

git clone https://github.com/LeoMartinezTAMUK/Ransomware_Family_Classification_ML.git

This Python program implements a Deep Neural Network (DNN) for classifying ransomware families using the Ransomware RISS Dataset. The dataset provides various features to analyze ransomware and classify it into different families.

Model Overview:

  • Language: Python 3.18
  • IDE: Spyder
  • Libraries Used:
    • numpy
    • pandas
    • matplotlib
    • seaborn
    • sklearn
    • keras

Dataset:

The dataset used for this project is the Ransomware RISS Dataset, which is accessible here. Due to its size, it is not included in this repository.

Implementation Details:

  • Data Loading: The dataset is loaded from a CSV file.
  • Data Splitting: The data is split into features (X) and target labels (y) for training and testing, with a 60/40 split.
  • Model Architecture:
    • Input Layer: Takes 30,967 binary features.
    • Hidden Layers:
      • 1st Layer: 256 neurons, ReLU activation
      • 2nd Layer: 128 neurons, ReLU activation
      • 3rd Layer: 64 neurons, ReLU activation
      • 4th Layer: 32 neurons, ReLU activation
    • Output Layer: 12 neurons (for 12 classes), Softmax activation

Training:

The model is compiled with the Adam optimizer and sparse categorical cross-entropy loss function. It is trained for 10 epochs with a batch size of 16. Early stopping is used with a patience of 6 to prevent overfitting.

Evaluation:

The model's performance is evaluated using the classification report, confusion matrix, and Matthews correlation coefficient (MCC). A confusion matrix heatmap is generated to visualize the results.

Visualization:

The training and validation accuracy are plotted and saved as training_validation_accuracy_plot.png. The confusion matrix heatmap is saved as Heatmap_400dpi.png. DNN Training Graph Normalized Heatmap

Note:

  • This project was created in Spyder, and separators ('#%%') are used for code sectioning.
  • For further details and similar projects, please visit my GitHub Page.

Here is a brief explanation of the items:

  • src: folder that contains the source code python script (.py)
  • images: contains the generated DNN training graph and confusion matrix heatmap (dataset is provided as a link due to size)
  • dataset_info: provides more context regarding the dataset from the original authors of the dataset
  • README.md: contains most basic information about the project
  • LICENSE: contains license information in regards to the Github repository