email-phishing-detection-svm

Email Phishing Detection – SVM Model

This is the SVM model trained to classify emails as Phishing or Legitimate using text-based features.

🧠 Model Details

• Type: Scikit-learn LinearSVC wrapped with CalibratedClassifierCV
• Supports: .predict_proba() for confidence scoring
• Vectorizer: Trained on TF-IDF transformed email text
• Training Set: Combination of phishing + legitimate emails from 7 public datasets
• Input: Preprocessed email text
• Output:
  • 1 = Phishing
  • 0 = Legitimate

📊 Performance

• Accuracy: ~98.38%
• Trained and validated on a clean split (~82K samples)
• Supports calibrated probability outputs

🔗 Web Demo

This model powers the live app:
https://phishingdetection.net

🧪 Usage

You’ll need to:

1. Preprocess text (lowercase, lemmatize, remove stopwords)
2. Transform with the corresponding TF-IDF vectorizer
3. Load and predict using this model:

import joblib
model = joblib.load("svm_model.pkl")
vector = vectorizer.transform(["your email text here"])
prediction = model.predict(vector)
confidence = model.predict_proba(vector)

Author

Emre Otu

---

license: mit language:

• en tags:
• phishing
• cybersecurity
• tfidf
• email
• sklearn
• support-vector-machine
• vectorizer
• calibrated-classifier model-index:
• name: Email Phishing SVM results: []

---