Malware detection using machine learning / Nur Wahida Kausar Abd Rahman

After detecting malware, categorizing risky files is a crucial part of the malware investigation process. So far, a number of static and dynamic malware classification algorithms have been reported. This study shows how malware families may be classified using a deep learning-based malware detection (DLMD) strategy based on static methodologies. To categorize malware families, the proposed DLMD approach uses both byte and ASM files for feature engineering. Two distinct Deep Convolutional Neural Networks are used first to extract features from byte input (CNN). Then, utilizing a wrapper-based method and a Support Vector Machine (SVM) as a classifier, important and discriminative opcode features are discovered. The objective is to mix several feature spaces to produce a hybrid feature space that overcomes each feature space's shortcomings and thereby minimizes the likelihood of malware being undetected. Finally, a Multilayer Perceptron is trained to categorize all nine malware types using the hybrid feature space. The proposed DLMD approach, according to experimental results, gives a log-loss of 0.09 for ten independent runs. Furthermore, the suggested DLMD approach's performance is compared to that of other classifiers, demonstrating its efficacy in detecting malware