This repository contains code for laser injection attack detection.
Clone this repository to your workspace using the following command.
git clone https://github.com/hashim19/Laser_Injection_Attack_Identification.git
Run the following command to go inside the parent directory of the repository.
cd Laser_Injection_Attack_Identification
Follow this tutorial to create a virtual environment in the parent directory of this repository
Activate the virtual environment using the following commands.
source *environment name* /bin/activate
Once, the environment is activated, run the following command in the parent directory of this repository to install the required libraries.
pip3 install -r requiremnts.txt
The dataset used to test the algorithms mentioned below is available here.
Download the dataset to a local directory.
Extract 4 types of features and save them in npy format in the audio_features directory. This file can also train an SVM model or a GMM model. Toggle train_SVM or train_GMM to True to train the SVM or GMM model respectively. The following types of features can be extracted.
- [CQCC] (https://ieeexplore.ieee.org/document/8659537)
- [DWT] (https://pywavelets.readthedocs.io/en/latest/)
- [LFCC] (https://spafe.readthedocs.io/en/latest/features/lfcc.html)
- [MFCC] (https://librosa.org/doc/0.10.1/index.html)
The following variables need to be set before the features can be extracted,
- audio_dir: Provide the path to the parent directory of the dataset Here.
- train_GMM: False if only extracting features
- train_SVM: False if only extracting features.
After configuring these variables, you can run the following command to extract features,
python3 extract_features.py
if train_GMM or train_SVM is true, a GMM or SVM model will be trained respectively and saved to the models directory.
This file generates the scores of the test audio using the trained models and saves the scores in the audio_features directory. Path to the dataset parent directory and audio_features folder needs to be provided.
Run python3 generate_scores.py to generate scores of trained models on the test audio.