Cybercrime is one of the major digital threats of this century. In particular, ransomware attacks have significantly increased, resulting in global damage costs of tens of billion dollars. In this paper, we train and test different Machine Learning and Deep Learning models for malware detection, malware classification and ransomware detection. We introduce a novel and flexible solution that combines two optimized models for malware and ransomware detection. Our results demonstrate some improvements both in terms of detection performances and flexibility. In particular, our combined models pave the way for easier future enhancements using specialized and thus interchangeable detection modules.
翻译:网络犯罪是本世纪的主要数字威胁之一。 特别是,赎金软件袭击大幅增加,导致全球损失成本达数百亿美元。 在本论文中,我们培训和测试了不同的机器学习和深学习模式,用于恶意软件检测、恶意软件分类和赎金软件检测。我们引入了一种新颖而灵活的解决方案,将两种最佳的恶意软件和赎金软件检测模式结合起来。我们的结果显示,在检测性能和灵活性两方面都取得了一些改进。特别是,我们的综合模型为利用专门和可互换的检测模块来更方便地改进未来工作铺平了道路。