【作者】 韩冰；

【导师】 刘吉强；

【作者基本信息】 北京交通大学 ， 信息安全， 2019， 硕士

【摘要】 物联网已经成为智慧城市、智慧小区等关系国计民生应用的重要组成部分。近年来,越来越多的物联网设备接入网络空间,包括网络摄像头、打印机、路由器、工业控制设备、智能家电等等,这些设备的安全问题引起了工业界和学术界的关注。物联网设备和设备地理位置信息对于保障网络空间安全起着关键作用。然而,目前的商业地理位置数据库只能提供粗粒度位置信息,定位到国家或者城市级别,无法满足细粒度物联网设备识别和定位的需求。本篇论文提出了大规模物联网设备的识别和定位算法,提高设备识别精确度和定位精度。本篇论文,首先实现了基于机器学习的物联网设备指纹生成算法。基于应用层数据报文内容,论文提取文本信息,结合通用的机器学习算法,生成物联网设备指纹,提高网络空间物联网设备识别精确率。其次,论文提出了基于测量的定位算法。具体来说,论文通过命名实体识别和正则表达式,提取可靠地理位置信息,生成大量的被动地标信息,作为测量的锚节点。论文基于欧几里得距离将地标节点分簇,获得高质量的地标信息,结合时延信息、网络拓扑结构和路由信息,计算目标物联网设备的经纬度信息,提高设备定位的精度。此外,论文通过调用谷歌地图接口实现图形化展示物联网设备的位置信息。本篇论文,通过搭建原型系统来验证算法的可行性。实验结果表明,物联网设备的识别达到了98%的识别精度和98%的覆盖率。在公开数据集合(近10000万数据报文)上,物联网设备指纹发现552万个物联网设备。在设备定位方面,论文提取了6万个具有经纬度信息的地标信息,定位精度控制在100公里以内。更多还原

【Abstract】 The Internet of Things has become an important part of smart cities,smart communities and other applications related to national economy and people’s livelihood.In recent years,more and more IoT devices have access to network space,including web cameras,printers,routers,industrial control devices,and smart home appliances.The security issues of IoT devices have drawn the attention of industry and academia.Geographic location information for IoT devices and devices plays a key role in ensuring cyberspace security.However,the current commercial geographic location database can only provide coarse-grained location information,which can only be located at the national or city level,and cannot meet the needs of identifying and locating fine-grained IoT devices.This paper proposes a context-based IoT device identification and localization algorithm to improve the accuracy of device identification and positioning.In this paper,a fingerprint generation algorithm for IoT devices based on machine learning is implemented.Based on the application layer data message content,the paper extracts the text information and combines with the general machine learning algorithm to generate the fingerprint of IoT devices and improve the identification accuracy of IoT devices in the network space.Secondly,this paper proposes a measurement-based localization algorithm.Specifically,this paper extracts reliable geographical location information by naming entity recognition and regular expression,and generates a large amount of passive landmark information as the anchor node for measurement.Based on the Euclidean distance,this paper clusters the landmark nodes to obtain high-quality landmark information,and combines the delay information,network topology and routing information to calculate the latitude and longitude information of the target IoT device and improve its accuracy.Device positioning.In addition,this article graphically displays the location information of IoT devices by calling the Google Maps interface.In this paper,the feasibility of the algorithm is verified by building a prototype system.The experimental results show that the identification accuracy and coverage rate of the IoT devices reach 98%.On public data collection(nearly 100 million pieces of data information),IoT devices fingerprints found 5.52 million IoT devices.In terms of equipment positioning,this paper extracts 60,000 landmark information with latitude and longitude information,and the positioning accuracy is controlled within 100 kilometers.更多还原