A Neural Network Approach For Malware Classification

Malicious program is commonly known as malware, which is purposely designed to cause problems in networks and harm in a variety of ways. Malware’s primary goal is to disrupt normal execution or to obtain unauthorized access. Malware is a generic term for variants of malicious software such as spyware, ransomware, and viruses. Malware application disguises itself like a standard program. The malware developer develops different variants of malware using obfuscation and compression strategies. Such concealment methods make detection and categorization of malware incredibly difficult. In order to automatically detect malware problems, many different machine learning algorithms have been deployed in last few years. This paper proposes an artificial neural network architecture for classifying malware variants. This model consists of four primary phases: data collection and pre-processing, designing of neural network model, training, and evaluation. The proposed technique is validated using the Microsoft BIG 2015 dataset. The experimental findings reveal that the proposed model effectively classifies malware with high precision and accuracy, as mentioned to different models in the literature. The proposed model shows good accuracy on training and testing data with 90.07% and 90.80%, respectively.