blizet/Llama-Phishing-Finetune

Phishing Detection Model - Fine-Tuned LLaMA-2

This repository contains a fine-tuned version of the LLaMA-2-7B model for phishing email detection. The model analyzes emails for phishing attempts, urgency, grammatical errors, and suspicious elements, providing detailed insights.

Model Details

• Base Model: NousResearch/Llama-2-7b-chat-hf
• Fine-Tuned Dataset: Custom phishing dataset (25,000 samples)
• Framework: Transformers, PEFT, TRL
• Quantization: 4-bit (QLoRA)

Features

1. Phishing Detection: Identifies suspicious elements in emails.
2. Context Analysis: Evaluates urgency, grammar, sender details, and more.
3. Detailed Responses: Outputs structured feedback for 15 parameters.

Example Usage

from transformers import pipeline

# Load the pipeline
pipe = pipeline("text-generation", model="blizet/Llama-Phishing-Finetune")

# Analyze an email
email_content = "Dear Customer, your account has been flagged. Click here to verify: http://example.com"
result = pipe(f"<s>[INST] Analyze this email for phishing: {email_content} [/INST]")
print(result[0]["generated_text"])