blizet/entbappy

Model Card for Phishing Email Detection Model

This model has been fine-tuned for phishing email detection, leveraging the NousResearch/Llama-2-7b-chat-hf base model and trained on the blizet/Phishing_Email dataset. The model is designed for text classification tasks, specifically to differentiate phishing emails from legitimate ones.

Model Details

Model Description

This model is a fine-tuned version of NousResearch's Llama-2-7b-chat-hf, specialized for phishing email detection. It has been trained on a dataset containing labeled phishing and legitimate email texts, enabling it to identify phishing attempts with high accuracy.

• Model type: Text Classification
• Language(s): English
• License: MIT
• Finetuned from model: NousResearch/Llama-2-7b-chat-hf

Uses

Direct Use

This model can be used directly to classify emails as phishing or legitimate. It can be integrated into email filtering systems, security applications, or used for research and educational purposes.

Out-of-Scope Use

The model is not designed for non-English emails, multi-modal inputs (e.g., image attachments), or advanced phishing detection in adversarial settings without further fine-tuning.

Bias, Risks, and Limitations

Bias

The model might show biases based on the dataset it was trained on. If the training data is unbalanced or skewed, the model might underperform on underrepresented email types or categories.

Risks

False positives may lead to legitimate emails being flagged as phishing, potentially causing disruptions. False negatives may result in phishing emails bypassing detection mechanisms.

Limitations

• Limited to English language emails.
• May not generalize well to entirely novel phishing strategies not present in the training data.
• Performance may degrade on noisy or poorly formatted text.

Recommendations

Users should validate the model’s performance on their specific data before deployment and periodically retrain the model with updated datasets.