Windows Malware Detection Using Quantum Long Short Term Memory

Abstract

The ongoing advancements have inadvertently led to a significant surge in malware attacks targeting the Windows platform. Malware detection is difficult due to the development of sophisticated and complex cyber-attacks as well as modern malware becoming highly dangerous. In this paper, the Deep Neural Network (DNN) namely Quantum Long Short Term Memory (QLSTM) is proposed for effective detection of malware. The classification of malware is enhanced by using the LASSO Regression based Feature Selection (LRFS). The Windows Portable Executable (PE) malware dataset is used to analyze the proposed LRFS-QLSTM. The proposed LRFS-QLSTM is analyzed using accuracy, sensitivity, F1-measure and recall. The existing approaches such as progressive deep unsupervised malware classification namely PROUD-MAL, Deep Learning (DL)-based Convolutional Neural Network (CNN), attention-based Multi-View DL (M-Attn) and LGBM are used to compare the LRFS-QLSTM. The F1-score of LRFS-QLSTM for Windows PE malware dataset is 99.36% which is high when compared to the PROUD-MAL, DL-CNN, M-Attn and LGBM.