PrivacyGuardian

PrivacyGuardian

**PrivacyGuardian** is a Security AI Agent that analyzes PDF documents for potential sensitive data leaks and information security risks through the **SenecaLLM** artificial intelligence model, which is developed specifically for cybersecurity. The tool extracts text from PDF files (either local files or via a web URL) and then uses advanced LLM models to provide a detailed analysis of the document’s content. It detects confidential information such as passwords, tokens, API keys, personally identifiable information (PII), and hard-coded credentials, and evaluates whether sharing the document with third parties may pose any risk. It provides users with the analysis results along with recommendations.

Features

• **PDF Analysis:** Supports both local PDF files and PDF documents accessed via web URLs.

• **LLM Integration:** Leverages Seneca LLM models for security analysis. Users can choose between:

  1. AlicanKiraz0/Seneca-x-DeepSeek-R1-Distill-Qwen-32B-v1.3-Safe
  2. AlicanKiraz0/SenecaLLM-x-Llama3.1-8B

• **Quantization Options:** Offers optional 4-bit or 8-bit quantization (using [bitsandbytes](https://github.com/TimDettmers/bitsandbytes)) to optimize model performance.

Demo Analysis Result:


Requirements

• Python 3.7+
• PyPDF2
• requests
• transformers
• langchain-huggingface
• accelerate
• bitsandbytes (required if using quantization)

Installation

You can install the required packages using pip:

pip install PyPDF2 requests transformers langchain-huggingface accelerate bitsandbytes