Deep Learning-Based Malware Analysis to Strengthen an Antivirus System

Malware detection remains a paramount challenge in cyber security due to malware's evolving nature and limitations in traditional detection techniques. This paper compares two deep learning models, an LSTM network and a Feed-Forward Neural Network (FFNN), for malware classification. The dataset comprises various features extracted from software samples, which are preprocessed and fed into the models for training and evaluation. The FFNN shows an accuracy of 99.96%, but the LSTM model, leveraging its capability to learn temporal dependencies, achieved an impressive accuracy of 99.989% on the first test. After changing the LSTM hyperparameters, 99.996% accuracy is achieved on the test set with performance improvement to 3 % compared to FFNN. This study highlights the potential of deep learning approaches in enhancing malware detection systems and provides insights into their implementation and performance.