网站指纹识别技术通过分析流量特征判断用户访问的网站站点,能够有效监管TOR匿名网络的用户行为。现有的识别方法通常需要大规模的数据样本以获得高的识别准确率,且普遍存在概念漂移问题。针对以上问题,本文提出一种基于残差和协作对抗...网站指纹识别技术通过分析流量特征判断用户访问的网站站点,能够有效监管TOR匿名网络的用户行为。现有的识别方法通常需要大规模的数据样本以获得高的识别准确率,且普遍存在概念漂移问题。针对以上问题,本文提出一种基于残差和协作对抗网络(Residual network and Collaborative and Adversarial Network,Re s-CAN)的网站指纹识别模型。该模型使用残差网络(Residual network)作为特征提取器以减少网络的优化难度。同时,将协作对抗网络(Collaborative and Adversarial Network,CAN)应用于网站指纹识别问题,使得特征提取器同时学习领域相关和领域无关特征,实现源域与目标域的特征空间对齐。实验结果表明,本文提出的方法在小样本环境下网站指纹识别准确率达到91.2%,优于现有的利用对抗领域自适应网络(Domain-Adversarial Neural Networks,DANN)迁移学习方法,且抗概念漂移能力较高。展开更多
网站指纹攻击利用统计方法来确定用户正在访问的网站,侵犯用户隐私,给互联网的安全和隐私带来了巨大挑战。网站指纹攻击首先收集用户访问不同网站时的数据,然后使用机器学习等方法处理数据,识别网站。之前的研究主要集中于传统的基于网...网站指纹攻击利用统计方法来确定用户正在访问的网站,侵犯用户隐私,给互联网的安全和隐私带来了巨大挑战。网站指纹攻击首先收集用户访问不同网站时的数据,然后使用机器学习等方法处理数据,识别网站。之前的研究主要集中于传统的基于网络流量的网站指纹攻击,本文重点介绍了较新出现的基于主机侧信道的网站指纹攻击,并讨论了这两类攻击的流程、指纹特征、威胁模型、分类方法、评价指标和防御研究。文章最后展望了网站指纹攻击的未来研究方向,尤其是新式的基于主机侧信道数据的网站指纹攻击当前存在的问题及未来发展方向。Website fingerprinting attacks utilize statistical methods to identify which websites a user is visiting, thereby infringing on user privacy and posing significant challenges to internet security and privacy. These attacks first collect data generated when a user visits different websites and then use methods such as machine learning to process the data and identify the websites. Previous research has mainly focused on traditional network traffic-based website fingerprinting attacks. This paper highlights the more recently emerged host-side channel-based website fingerprinting attacks and discusses the process, fingerprinting characteristics, threat models, classification methods, evaluation metrics, and defense measures of these two types of attacks. Finally, this paper provides an outlook on the future research directions of website fingerprinting attacks, particularly addressing the current issues and future development of host-side channel-based website fingerprinting attacks.展开更多
文摘网站指纹识别技术通过分析流量特征判断用户访问的网站站点,能够有效监管TOR匿名网络的用户行为。现有的识别方法通常需要大规模的数据样本以获得高的识别准确率,且普遍存在概念漂移问题。针对以上问题,本文提出一种基于残差和协作对抗网络(Residual network and Collaborative and Adversarial Network,Re s-CAN)的网站指纹识别模型。该模型使用残差网络(Residual network)作为特征提取器以减少网络的优化难度。同时,将协作对抗网络(Collaborative and Adversarial Network,CAN)应用于网站指纹识别问题,使得特征提取器同时学习领域相关和领域无关特征,实现源域与目标域的特征空间对齐。实验结果表明,本文提出的方法在小样本环境下网站指纹识别准确率达到91.2%,优于现有的利用对抗领域自适应网络(Domain-Adversarial Neural Networks,DANN)迁移学习方法,且抗概念漂移能力较高。
文摘网站指纹攻击利用统计方法来确定用户正在访问的网站,侵犯用户隐私,给互联网的安全和隐私带来了巨大挑战。网站指纹攻击首先收集用户访问不同网站时的数据,然后使用机器学习等方法处理数据,识别网站。之前的研究主要集中于传统的基于网络流量的网站指纹攻击,本文重点介绍了较新出现的基于主机侧信道的网站指纹攻击,并讨论了这两类攻击的流程、指纹特征、威胁模型、分类方法、评价指标和防御研究。文章最后展望了网站指纹攻击的未来研究方向,尤其是新式的基于主机侧信道数据的网站指纹攻击当前存在的问题及未来发展方向。Website fingerprinting attacks utilize statistical methods to identify which websites a user is visiting, thereby infringing on user privacy and posing significant challenges to internet security and privacy. These attacks first collect data generated when a user visits different websites and then use methods such as machine learning to process the data and identify the websites. Previous research has mainly focused on traditional network traffic-based website fingerprinting attacks. This paper highlights the more recently emerged host-side channel-based website fingerprinting attacks and discusses the process, fingerprinting characteristics, threat models, classification methods, evaluation metrics, and defense measures of these two types of attacks. Finally, this paper provides an outlook on the future research directions of website fingerprinting attacks, particularly addressing the current issues and future development of host-side channel-based website fingerprinting attacks.
