Deep Learning Malware Detection Using Auto Encoder

Abstract

Nowadays, in the arena of malware detection, the growing constraints of traditional detection methods, laterally with the improving precision of detection methods based on artificial intelligence algorithms, are moving research findings in this zone in favour of the latter. As a result, we offer a novel. This work includes a malware detection model. In a deep learning model, this model combines a grey-scale picture representation of malware with an autoencoder network, examines the viability of the grey-scale image approach malicious software based on the autoencoder reconstruction error, and employs the dimensionality.The auto encoder’s reduction characteristics are used to distinguish malware from benign software. Using the suggested detection model, the proposed detection model attained an accuracy of 96% and a steady F-score of about 96%. We collected an Android-side dataset that outperformed certain classic machine learning detection techniques.Malware detection, autoencoders, malware images, mobile application security.