A Hybrid DL-Based Framework to Classify Malware using Mexican Hat Wavelet Function
Detecting and categorizing malware represents a substantial and demanding undertaking within the
realm of information security and various other computer-related domains. Millions of malicious
files are detected annually. The high volume is largely due to malware authors using mutations to
evade detection, Malware variants are constantly evolving through the use of advanced obfuscation
and packing methods, making detection and classification increasingly difficult. In order to
efficiently examine and categorize a substantial volume of files, it becomes imperative to group
them and ascertain their behavioral characteristics to classify them effectively. In recent, most
malware classification techniques have been based on machine learning or deep learning models.
These models work with the train and test. The models are trained with the features, for instance,
opcode sequence, API calls, signature, etc. Recently, many deep learning techniques have been
proposed for Alex Net Network, Resnet-50 Network, and Hybrid (AlexNet-Resnet-50). These
models work well in terms of accuracy, Sensitivity, and so forth. However, these models are
complex in nature and need high computational power. In order to adequately confront the
difficulty presented by emerging malware variations, it becomes essential to employ alternative
approaches, as conventional artificial intelligence and machine learning algorithms are no longer
capable of identifying all intricate and constantly changing variants. A promising solution is deep
learning, which differs from traditional machine learning. This study proposes a Mexican hat
wavelet function that classifies malware variants through a hybrid deep learning model in this
approach, malware samples undergo conversion into grayscale images before being fed into the
DL system. Following the image acquisition section, the proposed method employs the convolution
layers of the hybrid architecture to extract high-level malware features from the malware images
with cloud-based architecture to decrease the computational intricacy, and neural network
complexity to achieve higher accuracy. Upon subjecting the proposed method to testing using the
MALIMG dataset, an accuracy of 99% was achieved. Similarly, when applied to the MALEVIS
dataset, an accuracy of 97.12% was attained, outperforming the majority of machine learning-based
methods employed for malware detection
A Hybrid DL-Based Framework to Classify Malware using Mexican Hat Wavelet Function
Detecting and categorizing malware represents a substantial and demanding undertaking within the
realm of information security and various other computer-related domains. Millions of malicious
files are detected annually. The high volume is largely due to malware authors using mutations to
evade detection, Malware variants are constantly evolving through the use of advanced obfuscation
and packing methods, making detection and classification increasingly difficult. In order to
efficiently examine and categorize a substantial volume of files, it becomes imperative to group
them and ascertain their behavioral characteristics to classify them effectively. In recent, most
malware classification techniques have been based on machine learning or deep learning models.
These models work with the train and test. The models are trained with the features, for instance,
opcode sequence, API calls, signature, etc. Recently, many deep learning techniques have been
proposed for Alex Net Network, Resnet-50 Network, and Hybrid (AlexNet-Resnet-50). These
models work well in terms of accuracy, Sensitivity, and so forth. However, these models are
complex in nature and need high computational power. In order to adequately confront the
difficulty presented by emerging malware variations, it becomes essential to employ alternative
approaches, as conventional artificial intelligence and machine learning algorithms are no longer
capable of identifying all intricate and constantly changing variants. A promising solution is deep
learning, which differs from traditional machine learning. This study proposes a Mexican hat
wavelet function that classifies malware variants through a hybrid deep learning model in this
approach, malware samples undergo conversion into grayscale images before being fed into the
DL system. Following the image acquisition section, the proposed method employs the convolution
layers of the hybrid architecture to extract high-level malware features from the malware images
with cloud-based architecture to decrease the computational intricacy, and neural network
complexity to achieve higher accuracy. Upon subjecting the proposed method to testing using the
MALIMG dataset, an accuracy of 99% was achieved. Similarly, when applied to the MALEVIS
dataset, an accuracy of 97.12% was attained, outperforming the majority of machine learning-based
methods employed for malware detection