vicgc / ICT2202-DF-Assignment_One

Automated Media Analysis Modules for Sleuthkit Autopsy

Team Trying Hardest

Automated Media Analysis Modules for Sleuthkit Autopsy

The project aims to create media analysis plug-ins for the Autopsy - Digital Forensics platform. With the use of open-source machine learning engines and technologies, we have been able to create a suite of plug-ins on the Python programming language that integrates and extends the functionality of Autopsy.

Image Analyser

• Object Detection and Identification for still images.
• Optional Facial Recognition.
• Based on ImageAI

Video Analyser

• Object detection and identification on video files.
• Based on ImageAI

AudioToText

• Audio to text transcription.
• Based on DeepSpeech

Project Folder Structure

.
├── DeepSpeech/         # AudioToText plugin
├── VideoAI/            # Video Analyser plugin
│   └── models/         # imageAI pretrained model
├── imageAI/            # Image Analyser plugin
│   └── models/         # imageAI pretrained model
├── sample/             # Sample test data
└── README.md

Platform Requirements

Tested and developed on

• Microsoft Windows 10
• Autopsy 4.19.1
• Java 1.8 64-bit
• Python 3.7

Key validated Python libraries & versions

Versions and dependencies are listed within the installation guide

• Tensorflow 2.4.0
• ImageAI 2.1.6
• face_recognition 1.3.0
• DeepSpeech 0.9.3

Validated Models

• ImageAI v1.0
  • yolo-tiny.h5
• DeepSpeech v0.9.3
  • deepspeech-0.9.3-models.pbmm
  • deepspeech-0.9.3-models.scorer

Installation Manual

Prerequisites - Installing the support files

Due to potential conflicts in dependency versions it is highly recommended that you install the following files in the provided sequence.

1. The basics

   • Ensure you have Python 3.7 installed! Download here
   • Python Dependencies

     pip install --upgrade pip
     

     pip install wheel
     

2. Image Analyser / Video Analyser

   • Python Dependencies

     pip install tensorflow==2.4.0 keras==2.4.3 numpy==1.19.3 pillow==7.0.0 scipy==1.4.1 h5py==2.10.0 matplotlib==3.3.2 opencv-python keras-resnet==0.2.0 imageai
     

   • Others
     • The ImageAI model has been pre-bundled
     • If you wish to replace the bundled model you can do so at the plugin sub-directory ./imageAI/models/yolo-tiny.h5 and ./VideoAI/models/yolo-tiny.h5

3. Facial Recognition for Image Analyser

   • Python Dependencies

     pip install face_recognition
     

   • If the install fails on Windows when installing dlib, please install
     • CMake
     • Microsoft C++ (MSVC) compiler toolset

4. AudioToText

   • Python Dependencies

     pip install deepspeech pydub webrtcvad
     

   • Others
     • FFMPEG to support conversion from .mp3 to .wav for processing
     • DeepSpeech Model to be downloaded and selected when prompted
       • deepspeech-0.9.3-models.pbmm
       • deepspeech-0.9.3-models.scorer

Installation - Setting up the plugin on Autopsy

1. Either download via

   • The release bundle Latest Release
   • Or any other method e.g git

2. Move the following folders into the Autopsy Python plugin directory

   • DeepSpeech
   • VideoAI
   • imageAI

   Plugin directory on Windows is %appdata%\autopsy\python_modules\

   Your directory tree should now look like

   python_modules\
   ├── DeepSpeech\         # AudioToText plugin
   ├── VideoAI\            # Video Analyser plugin
   └── imageAI\            # Image Analyser plugin
   

3. Ensure DeepSpeech Models have been downloaded as per Prerequisites - 4. AudioToText

4. Launch Autopsy!

User Guide - How to use the plugins

A Walkthrough of the Autopsy Module Suite