KRAMS: Keystroke Recognition using Augmented Mel-Spectrograms

A command-line interface application that uses SpecAugmented Mel-spectrograms to train a CoAtNet deep-learning model that classifies keystrokes in an attack recording.

The application takes a directory containing training recordings, an attack recording, and the number of keystrokes in that recording as input.

The method used is inspired by this paper.

WARNING: It is highly recommended to use a GPU to run this application. The training process is computationally expensive and will take a long time on a CPU. Services such as Google Colab can be used to run the application on a GPU for free.

Installation

To get started, follow these simple steps:

Clone repository:

git clone https://github.com/moahmed0987/KRAMS.git

Set up virtual environment:

• Windows:

  cd KRAMS
  python -m venv venv
  .\venv\Scripts\activate

• MacOS/Linux:

  cd KRAMS
  python3 -m venv venv
  source venv/bin/activate

Install dependencies:

pip install -r requirements.txt

Usage

Simply run the following command replacing the placeholders with your specified file paths:

python KRAMS.py <TRAINING_RECORDINGS_DIR> <ATTACK_RECORDING_PATH> <N_KEYSTROKES_IN_ATTACK>

TRAINING_RECORDINGS_DIR: Path to the directory containing the training recordings.

ATTACK_RECORDING_PATH: Path to the recording to be attacked.

N_KEYSTROKES_IN_ATTACK: Number of keystrokes in the attack recording.