A Hybrid Deep Learning Malware Detection Model with The Fusion of VGG-16 and ResNet-50 Architectures

Recent advancements in computer technology enable people to spend most of the time with the Internet for various purpose. This helps the cybercriminals to commit online crime with a little physical effort than in the real world. The unwanted softwares known as malicious software (malware) are the primary weapon of cybercriminals to conduct cyberattacks. Several approaches have been deployed till date and among all the ML techniques are better than others provided the models are trained with realistic dataset. The Deep Learning (DL) is the future of ML with its success almost in every applications. This research propose a revolutionary architecture grounded in DL that is capable of classifying malware types using a hybrid approaches. The proposed one is a hybrid DL combines VGG-16 and ResNet 50 model that significantly improves accuracy. The performance is verified through three standard datasets: The Microsoft BIG 2015, Malimg, and MaleVis. The experimental results demonstrate that the proposed method outperforms the most recent techniques with a highest accuracy of 99.28% with the Malimg dataset.