Charleswyt / audio_steganalysis_ml

Audio steganalysis based on traditional handcrafted features design.

Audio Steganalysis with Handcrafted Features Design (Statistical Machine Learning)

Audio steganalysis based on the agorithms of statisctical machine learning.
@ Author: Yuntao Wang (Charles_wyt)
@ Email: [e-mail]
Hope we can have a happy communication.

This project is the implementation of our recent work for audio steganalysis based on statistical machine learning, and you can also design your own algoritm through this platform.

Files

ID	File	Function
1	application	audio steganalysis and steganographied find
2	batch_script	all batch scripts for feature extraction, training, test and so on
3	data_processing	tools which are used for QMDCT coefficients extraction and dataset build
4	feature_extract	the scripts for feature extraction (ADOTP, MDI2, I2C, D2MA, JPBC, Co-Occurance)
5	plot	scripts for figure plot
6	train_test	training, validation and test via svm and ensemble classifier
7	utils	some basic tools such as get files name and get files list

How to use

Single task

Separation

1. Run setup.m and complete environmental configuration.
2. For QMDCT extraction, run data_processing/batch_script/QMDCT_extraction_batch1.bat or QMDCT_extraction_batch2.bat.
3. For feature extraction, run matlab scripts of batch_script/feature_extraction_batch.m.
4. For training and validation, run matlab scripts of train_test/ensemble_classifier/training_emsemble.m or train_test/svm_classifier/training_svm.m.
5. For test, run matlab scripts of train_test/ensemble_classifier/test_ensemble.m or train_test/svm_classifier/test_svm.m.

Integration

1. Run setup.m and complete environmental configuration.
2. Run run_script_for_training1.m or run_script_for_training2.m for training and validation.
3. Run run_script_for_test.m for test.

Multiple tasks

1. Run setup.m and complete environmental configuration.
2. Run run_script_for_experiments.m or run_script_for_experiments1.m, and all results are writtten into a text file.