Intelligent Framework for Malware Detection with Convolutional Neural Network
Intelligent Framework for Malware Detection with Convolutional Neural Network
复制标题
使用卷积神经网络进行恶意软件检测的智能框架
DOI：
10.1145/3320326.3320333
复制
发表时间：
2019
期刊：
影响因子：
--
通讯作者：
D. Alghazzawi
中科院分区：
文献类型：
--
作者：
Youness Mourtaji;M. Bouhorma;D. Alghazzawi
在本文中，我们提出了一个用于恶意软件分类的深度学习框架。近年来，恶意软件的数量激增，这给金融机构、机构和个人带来了极大的安全机会。为了对抗恶意软件的扩散，需要新的技术来快速感知和分类恶意软件样本，以便能够分析它们的行为。机器学习方法因分类恶意软件而闻名，但现代小工具中最大的部分是使用机器学习算法(例如，支持向量机)来获取恶意软件分类策略的知识。目前，卷积神经网络(CNN)作为一种深度知识获取方法，与传统的知识获取算法相比，已经被证明具有优越的性能，特别是在包括图像分类在内的任务中。受这一成果的影响，我们建议使用基于CNN的体系结构来对恶意软件样本进行分类。我们将恶意软件二进制文件转换为灰度图像，最后，我们训练CNN网络进行分类。在硬恶意软件分类数据集Malimg和微软恶意软件上的实验表明，我们的技术取得了比现代整体性能更高的性能。
In this paper, we propose a deep learning framework for malware classification. There was a big boom within the quantity of malware in current years which poses an extreme safety chance to financial establishments, agencies, and people. In order to fight the proliferation of malware, new techniques are essential to quickly perceive and classify malware samples so that their behavior can be analyzed. Machine learning methods are becoming famous for classifying malware, but, the maximum of the modern gadget gaining knowledge of strategies for malware classification use machine learning algorithms (e.g., SVM). Currently, Convolutional Neural Networks (CNN), a deep getting to know approach, have proven advanced performance in comparison to traditional getting to know algorithms, particularly in duties which include image classification. Influenced by way of this achievement, we recommend a CNN-based architecture to classify malware samples. We convert malware binaries to a grayscale image, and at the end, we train a CNN network for classification. Experiments on hard malware classification datasets, Malimg, and Microsoft malware, reveal that our technique achieves higher than the modern-day overall performance.