Objective: Newborn malware increase significantly in recent years, becoming more dangerous for many applications. So, researchers are focusing more on solutions that serve the defense of new malwares trends and variance, especially zero-day malware attacks. The prime goal of our proposition is to reach a high security level by defending against malware attacks effectively using advanced techniques. Methods: In this paper, we propose an Intelligent Cybersecurity Framework specialized on malware attacks in a layered architecture. After receiving the unknown malware, the Framework Core layer use malware visualization technique to process unknown samples of the malicious software. Then, we classify malware samples into their families using: K-Nearest Neighbor, Decision Tree and Random Forest algorithms. Classification results are given in the last layer, and based on a Malware Behavior Database we are able to warn users by giving them a detail report on the malicious behavior of the given malware family. The proposed Intelligent Cybersecurity Framework is implemented in a graphic user interface easy to use. Results: Comparing machine learning classifiers, Random Forest algorithm gives best results in the classification task with a precision of 97,6%. Conclusion: However, we need to take into account results of the other classifiers for more reliability. Finally, obtained results are as efficient as fast that meets cybersecurity frameworks general requirements.
Malware classification is an important task in cybersecurity, where machine learning techniques have been widely used to automate the process of identifying and categorizing malicious software. In this paper, we propose a new approach to malware classification using Vision Transformers (ViT), a state-of-the-art deep learning architecture that has shown good results in computer vision field. Our proposed ViT-based model outperforms traditional Convolutional Neural Networks (CNNs) in terms of accuracy and robustness, as it can capture long-range dependencies in the input data without relying on hand-crafted features. We evaluate our proposed model on Malimg database, and demonstrate that it achieves state-of-the-art performance, outperforming the existing traditional approaches. Furthermore, we investigate the impact of different input representations, model configurations, and training strategies on the ViT-based model's performance. Our results show that the proposed ViT-based model offers several advantages over traditional CNN models, such as better performance on large-scale and complex datasets, higher interpretability, and scalability. However, the ViT-based model requires significantly more computational resources and longer training time. Our proposed approach offers a promising direction for malware classification using ViT-based models, which can be further improved by exploring different architectures, optimization techniques, and transfer learning strategies.