Unified Malware detection approach: Leveraging optimal features across diverse platforms Available to Purchase

One of the most frequent cyberattacks is malware, which is spreading throughout networks at an accelerating rate. Traffic including malware is always asymmetric compared to benign software traffic, which is always symmetric. Fortunately, malware can be identified and distinguished from legitimate activities using a number of artificial intelligence techniques. However, there hasn’t been enough attention paid to the issue of handling massive and high-dimensional data. This study presents a high-performance malware detection system that makes use of feature selection and machine learning techniques. Huge datasets used involve Two different malware groups for Android and Windows are used to detect malware and distinguish it from benign activities. The software sets are pre-processed, features are extracted and then feature selection is applied using Cuckoo Search to produce different feature-selected datasets. The four machine learning models: Random Forests, Gradient Boosting, Stochastic Gradient Descent, and Extra Trees are then trained using two different sets of data sets extracted from each system and defined by features. Afterwards, a number of performance indicators (accuracy, precision, recall, and F1 score) are used to assess the trained models. The results indicate that the Random Forests algorithm achieved the highest accuracy, reaching 99%, on the data of both systems, while the rest of the algorithms achieved varying accuracy.