A robust and reliable method for malware classification using hybrid features and gradient boosting

Abstract

The most significant problem facing software security firms in the current cybersecurity system is the fact that new strains of malware keep on appearing. The achievement of successful detection of malicious software based on both static and dynamic data remains unmet despite the further realization of different approaches to deep learning and machine learning. Defects in encryption, packaging, and distribution, the discrepancies between the distribution of malware samples in families, make this issue even worse. To do actual malware research that utilizes deep learning methods, you will require data scientists of appropriate competence, large amounts of computer power (particularly GPUs) and time. In order to overcome such issues, this approach involves a hybrid approach to feature extraction that integrates wavelet and other statistical features based on benchmarked malware image sets. A classification model based on gradient-boosting can be used reliably to classify different types of malwares. Further, this paper examined the capability of the machine learning system to Classify in-distribution test samples well. This method is properly tested on the two benchmarked datasets MelVis and MALIMG to ensure its effectiveness. By joining these two sets of data it is possible to test the suggested technique using varied array of samples. The proposed approach outperforms the state-of-the-art techniques with an incredible accuracy rate of 100% using MALIMG and 98.10 and 98.40% using combined datasets. Its value extends far beyond cybersecurity in such spheres as sustaining IoT ecosystems to assist in improving networked devices and systems.