Efficient Image based Malware Classification Using a Modified VGG based deep Learning Model

Abstract

The rapid development of internet has led to an increase in attack methods and malware. Every year, anti-virus companies find millions of new variations of malware.  Several organizations created novel methods to protect persons from such scams.  Malware is increasing in frequency, variety, and sophistication. To stop this rise, novel malware detection approaches should be developed. Recent research has shown that deep learning is very effective at detecting malware in images. For malware detection, DL algorithms like modified VGG are used with an image-based malware dataset. The pre-processed images were used to train DL model first. The dataset is later segmented into training and testing data. For the experimental setting, the proposed model, MalNet successfully identified malware images. MalNet was then used to categorize malware images and was compared to other trained models. The suggested method produced very accurate and precise results.