Phishing-Detection

Phishing Detection Model using Logistic Regression and TF-IDF

This model is a phishing detection classifier built using TF-IDF for feature extraction and Logistic Regression for classification. It processes text data to identify phishing attempts with high accuracy.

Model Details

• Framework: Scikit-learn
• Feature Extraction: TF-IDF Vectorizer (top 5000 features)
• Algorithm: Logistic Regression
• Dataset: ealvaradob/phishing-dataset (combined_reduced subset)

Installation

Before using the model, ensure you have the necessary dependencies installed:

pip install scikit-learn
pip install -U "tensorflow-text==2.13.*"
pip install "tf-models-official==2.13.*"
pip uninstall -y pyarrow datasets
pip install pyarrow datasets

How to Use

Below is an example of how to train and evaluate the model:

from datasets import load_dataset
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.linear_model import LogisticRegression
from sklearn.model_selection import train_test_split
from sklearn.metrics import accuracy_score

# Load the dataset
dataset_reduced = load_dataset("ealvaradob/phishing-dataset", "combined_reduced", trust_remote_code=True)

# Convert to pandas DataFrame
df = dataset_reduced['train'].to_pandas()

# Extract text and labels
text = df['text'].values
labels = df['label'].values

# Split the data into train and test sets
train_text, test_text, train_labels, test_labels = train_test_split(
    text, labels, test_size=0.2, random_state=42
)

# Create and fit the TF-IDF vectorizer
vectorizer = TfidfVectorizer(max_features=5000)
vectorizer.fit(train_text)

# Transform the text data into numerical features
train_features = vectorizer.transform(train_text)
test_features = vectorizer.transform(test_text)

# Create and train the logistic regression model
model = LogisticRegression()
model.fit(train_features, train_labels)

# Make predictions on the test set
predictions = model.predict(test_features)

# Evaluate the model's accuracy
accuracy = accuracy_score(test_labels, predictions)
print(f'Accuracy: {accuracy}')

Results

• Accuracy: The model achieves an accuracy of {{accuracy}} on the test set.

Dataset

The dataset used for training and evaluation is the ealvaradob/phishing-dataset. It contains a variety of phishing and non-phishing samples labeled as 1 (phishing) and 0 (non-phishing).

Limitations and Future Work

• The model uses a simple Logistic Regression algorithm, which may not capture complex patterns in text as effectively as deep learning models.
• Future versions could incorporate advanced NLP techniques like BERT or transformer-based models.

License

This project is licensed under the MIT License. Feel free to use, modify, and distribute this model as per the terms of the license.

Acknowledgements

• Hugging Face Datasets
• Scikit-learn

---

license: apache-2.0