机器学习安全性问题及其防御技术研究综述.docVIP

机器学习安全性问题及其防御技术研究综述 李盼 赵文涛 刘强 崔建京 殷建平 国防科技大学计算机学院 X 关注成功！ 加关注后您将方便地在 我的关注中得到本文献的被引频次变化的通知！ 新浪微博 腾讯微博 人人网 开心网 豆瓣网 网易微博 摘????要： 机器学习已经成为当前计算机领域研究和应用最广泛的技术之一, 在图像处理、自然语言处理、网络安全等领域被广泛应用。然而, 一些机器学习算法和训练数据本身还面临着诸多安全威胁, 进而影响到基于机器学习的面部检测、恶意程序检测、自动驾驶汽车等实际应用系统的安全性。由目前已知的、针对SVM分类器、聚类、深度神经网络 (DNN) 等多种机器学习算法的安全威胁为出发点, 介绍了在机器学习的训练阶段和测试/推理阶段中出现的、基于对抗样本的投毒、逃逸、模仿、逆向等攻击和隐私泄露等问题, 归纳了针对机器学习的敌手模型及其安全评估机制, 总结了训练过程和测试过程中的若干防御技术和隐私保护技术, 最后展望了下一步机器学习安全研究的发展趋势。 关键词： 机器学习; 对抗样本; 安全威胁; 防御技术; 作者简介：李盼 (1992-) , 男, 河北邯郸人, 国防科技大学硕士研究生, 主要研究领域为机器学习, 网络安全。 作者简介：赵文涛 (1973-) , 男, 内蒙凉城人, 2008年于国防科技大学获得博士学位, 现为国防科技大学教授, 主要研究领域为网络空间安全态势感知。 作者简介：ZHAO Wentao, was born in 1973. He received the Ph.D. degree from National University of Defense Technology (NUDT) in 2008. Now he is a professor at NUDT. His research interest is cyberspace security situation awareness. Security Issues and Their Countermeasuring Techniques of Machine Learning : A Survey LI Pan ZHAO Wentao LIU Qiang CUI Jianjing YIN Jianping College of Computer, National University of Defense Technology; Abstract： Machine learning has already become one of the most widely used techniques in the field of Computer Science, and it has been widely applied in image processing, natural language processing, network security and other fields. However, there has been many security threats need to be overcome on current machine learning algorithms and training data set, which will affect the security of several practical applications such as facial detection, malware detection, automatic driving and etc. According to the known security threats, which aimed to a variety of machine learning algorithms such as the SVM classifier, clustering and deep neural networks, this work introduce the issues that happened in the training, testing/inference phase of machine learning, which include privacy leaking and attacks of poisoning, evasion, impersonate and inversion based on the adversarial samples, sum up the machine learning adversary model as well