Vulnerability Detection Framework

A vulnerability detection framework with a fine-tuned ConvNext model. It focuses on the Windows remote desktops and reads the IP addresses to check the login interfaces.

Examples

This project takes IP addresses as input, and scan the login interfaces of Windows remote desktops.

High vulnerability: Activated admin account and system update needed.

Low vulnerability: No ativated admin account or system update needed.

Medium vulnerability: No Activated admin account but system update needed.

Medium vulnerability: Activated admin account but no system update needed.

Base model

Fine-tuned ConvNext-1K, 384 resolution. ConvNeXt model The ConvNeXt model is introduced by Zhuang Liu, Hanzi Mao, Chao-Yuan Wu, Christoph Feichtenhofer, Trevor Darrell, and Saining Xie in A ConvNet for the 2020s. You can find more in their official repository ConvNeXt.

Usage

Change directory to project folder

cd Vulnerability-Detection-Framework

And call the main program

python main.py [-h] [--use_local] [--output_dir] input_ads

usage: main.py [-h] [--use_local] [--output_dir] input_ads
positional arguments:
  input_ads             input_ads the IP address or a directory of text file

options:
  -h, --help            show this help message and exit
  --use_local           load model from local
  --output_dir OUTPUT_DIR
                        directory of output text file

Input an IP address (ex. 0.0.0.0:1000) or a text file that stores multiple IP addresses.

Local Model Option

Download config file and model in directory /Vulnerability-Detection-Framework/model. If you want to apply my model locally, you can download here.

----- Vulnerability-Detection-Framework
    ----- model
        - config.JSON
        - model.safetensors
        - preprocessor_config.json
    ----- src
    - main.py

Note that this project only support custom model and feature extractor for now.

Catalog

• Add pre-trained processor option.
• Google Colab inference example.
• Fine-tuning code.
• Data collecting code (Shodan.io API).