0.1.0

ADFLER: Advanced Drone Flight Log Entity Recognizer

ADFLER is an open-source tool developed to perform named entity recognition on drone flight log data to support forensic investigations. It uses a fine-tuned ALBERT model to identify entities (like Event and NonEvent) in log messages and constructs a forensic timeline.

Features

• Forensic Timeline Construction: Merges and sorts logs from Android and iOS devices.
• Entity Recognition: Uses an ALBERT model to highlight key entities in log messages.
• Forensic Report Generation: Generates a PDF report with statistics and the highlighted timeline.

Installation

Prerequisites

1. Python 3.7+
2. wkhtmltopdf: Required for report generation.
   • Windows: Download and install from wkhtmltopdf.org. Default path: C:\Program Files\wkhtmltopdf\bin\wkhtmltopdf.exe.
   • Linux: sudo apt-get install wkhtmltopdf. Default path: /usr/bin/wkhtmltopdf.

Install ADFLER

You can install ADFLER directly from PyPI (coming soon) or from source.

From Source:

git clone https://github.com/DroneNLP/ADFLER.git
cd ADFLER
pip install .

Model

ADFLER uses a fine-tuned ALBERT model (swardiantara/ADFLER-albert-base-v2). By default, the tool will automatically download and cache the model from Hugging Face on its first run.

If you prefer to use a locally downloaded model or an offline environment, you can point to your local directory containing pytorch_model.bin using the --model argument.

Usage

ADFLER provides a Command Line Interface (CLI).

Structure:

adfler [arguments]

Arguments

Examples

1. Run Pipeline (Recommended):

adfler --evidence "./flight_logs" --output "./results/case_001"

2. Run Pipeline with Custom Model:

adfler --evidence "./flight_logs" --output "./results/case_001" --model "./local_model_dir"

Output Structure

Running adfler will create the following structure in your output directory:

output_dir/
├── raw_list.json             # List of detected log files
├── forensic_timeline.csv     # Merged and sorted timeline of all logs
├── ner_result.json           # Timeline with detected entities (JSON format)
├── statistics.json           # Counts of detected entity types
├── forensic_report_.html     # Intermediate HTML report
├── forensic_report_.pdf      # Final PDF Forensic Report
└── parsed/                   # Intermediate parsed CSVs
    ├── android/
    └── ios/

Configuration

You can optionally use a config.json file instead of passing long arguments.

Example config.json:

{
    "source_evidence": "./flight_logs",
    "output_dir": "./results/test_run",
    "model_dir": "./model",
    "wkhtml_path": {
        "windows": "C:\\Program Files\\wkhtmltopdf\\bin\\wkhtmltopdf.exe",
        "linux": "/usr/bin/wkhtmltopdf"
    },
    "use_cuda": true
}

Run with config:

adfler --config my_config.json

License

MIT