Android-Malware-Detection

Android Malware Detection using Deep Learning

Classification of android apps done based on pseudo-dynamic analysis of system API Call sequences.
Developed a Deep Convolutional Neural network (CNN) and a Recurrent Neural Network (LSTM) model.
Compared performance of these Deep Neural Network models with Naive Bayes Classifier.
Developed an autoencoder model and fed the compressed representation to the CNN model .

Files

File	Description
generate_dict.py	Generates global dictionary for storing mapping all distinct API Calls to numbers in the dataset and pickles the dictionary
load_dict.py	Loads the dictionary of API Calls
extract_all_features.py	Extracts all feature vectors (of size n X m X h) for 8 testcases in dataset and pickles them
extract_all_features_compressed.py	Extracts all feature vectors (of size n X h) for 600 apps in dataset and pickles them
data_reader.py	Loads all pickled feature vectors
uncompress.py	Uncompresses features to one-hot form
cnn2.py and cnn3.py	Different cnn architectures
lstm.py	lstm model
nb.py	Naive bayes model
ae.py	Stacked Autoencoder
ae_cnn.py	Training Cnn with Stacked autoencoder
ae_cnn_test.py	Testing Cnn with Stacked autoencoder