Detection of PE Malware Files using Machine L...

Detection of PE Malware Files using Machine Learning and Deep Learning Techniques

Author Name : Rakesh Kumar B Sahu

DOI: https://doi.org/10.56025/IJARESM.2022.1091269

ABSTRACT

Malware detection is a necessary step in the process of identifying threats so that the user's security can be preserved. Static analysis is also used in conjunction with this method, and it takes more time to extract features and feed them to the machine for testing. This paper intends to focus on deploying a model that could employ the principles of feature extraction and subsequently classify the PE Files as malware or benign. The collected features from the study are put into deep learning and machine learning algorithms, where they are processed through layers of neural networks to improve the model's overall architecture. Additionally, the model goes through the feature selection and pre-processing processes. The algorithms are then trained using the pre-processed data. The Kaggle and VirusShare are used as dataset repositories. The dataset is initially trained using three machine learning-based techniques, and then a stacking algorithm using a base classifier of random forest and AdaBoost and a Meta classifier of XGBoost. Gradient Boosting among the three classifiers gave 98 percent accuracy through this implementation; however the stacking technique produced the optimal accuracy of 99 percent. On the other hand, the application of CNN coupled with LSTM provided better accuracy than the implementation of CNN alone.

Keywords: Deep Learning, Kaggle, Machine Learning, PE Malware, VirusShare