Deployment of IoT cameras in an organization threatens security and privacy policies, and the classification of network traffic without using IP addresses and port numbers has been challenging. In this paper, we have designed, implemented and deployed a system called iCamInspector to classify network traffic arising from IoT camera in a mixed networking environment. We have collected a total of about 36GB of network traffic containing video data from three different types of applications (four online audio/video conferencing applications, two video sharing applications and six IoT camera from different manufacturers) in our IoT laboratory. We show that with the help of a limited number of flow-based features, iCamInspector achieves an average accuracy of more than 98% in a 10-fold cross-validation with a false rate of about 1.5% in testing phase of the system. A real deployment of our system in an unseen environment achieves a commendable performance of detecting IoT camera with an average detection probability higher than 0.9.
翻译:在本文中,我们设计、实施和部署了一个名为iCamInspecter的系统,对在混合网络环境中由IoT相机产生的网络流量进行分类,我们收集了总共约36GB的网络流量,其中包含来自我们IoT实验室三种不同应用的视频数据(4个在线音像会议应用程序、2个视频共享应用程序和6个不同制造商的IoT相机)。我们显示,在数量有限的流动特征的帮助下,iCamInspector在10倍交叉校验中平均达到98%以上,在系统测试阶段的误率约为1.5%。在不为人知的环境中实际部署我们的系统,在探测IoT相机方面取得了令人称道的成绩,平均探测概率高于0.9。