摘要
互联网规模的急剧增长导致IPv4地址资源逐渐耗尽,IPv6的大规模部署有效地解决了IPv4地址耗尽的问题.然而,IPv6地址空间庞大的特性为活跃地址的探测带来了巨大挑战.当前已有的活跃IPv6地址探测方法存在探测速度较慢、命中率偏低、探测范围受限等问题.针对这些挑战,本文提出了高效、快速、适用范围广的活跃地址探测系统6EDL.6EDL将地址探测分为无种子地址场景和有种子地址场景,并针对每种场景设计高效探测算法.在无种子地址场景下,本文提出了6EDL-N,使用神经网络挖掘BGP前缀信息与地址配置模式之间的潜在关系,实现了有种子区域到任一无种子区域的地址迁移,从而扩展了地址探测的边界.此外,本文引入了预探测机制,有效缓解了大规模地址探测中的资源浪费问题.实验表明,6EDL-N的命中率达到12.69%,覆盖度为21.97%,单位时间发现的活跃地址数(NPT)为233.09个/s.与现有工作相比,6EDL-N的命中率是其的8.13倍,NPT为14.94倍,覆盖度为1.84倍.在有种子地址场景下,本文提出基于生成对抗网络(GAN)的活跃地址探测方法6EDL-S,通过精细的种子地址分布规律学习,并采用环境反馈机制来缓解种子地址采样偏差,有效提升了命中率.实验表明,6EDL-S的命中率达到了25.91%,是已有方法的1.23~10.89倍.同时,NPT为466.72个/s,是已有方法的1.49~6.20倍.最终,经过持续探测,6EDL系统成功发现了29.77亿个活跃地址,包含5.66亿别名地址和24.11亿非别名地址,覆盖了125101个BGP前缀和40137个AS.本文构造的活跃IPv6地址集将有效支撑IPv6网络测量和安全分析等多种应用,进一步打开了IPv6网络研究的大门.
The rapid growth of the size of the Internet has led to the gradual depletion of IPv4 address resources,and the large-scale deployment of IPv6 has effectively solved the problem of IPv4 address exhaustion.However,the vast expanse of the IPv6 address space presents significant challenges for the detection of active IPv6 addresses.Existing methods for detecting active IPv6 addresses suffer from issues such as slow speed,low hit rates,and limited detection coverage.To address these challenges,we propose the active IPv6 address detection system 6EDL,which is efficient,fast,and broadly applicable.6EDL divides active IPv6 address detection into scenarios without seed addresses and with seed addresses(e.g.,the known active IPv6 addresses),and designs efficient active IPv6 address detection algorithms tailored to each scenario.In the scenario without seed addresses,we propose the 6EDL-N method.This method uncovers the latent relationship between BGP prefix information and address configuration patterns,enabling the migration of addresses from areas with seed addresses to any area without seed addresses,thereby extending the boundary of address detection.Additionally,we establish a pre-scanning mechanism that effectively mitigates resource waste in large-scale address detection.The experimental results show that in the scenario without seed addresses,6EDL-N achieves a hit rate of 12.69%,a coverage rate of 21.97%,and the number of active IPv6 addresses discovered per unit time(NPT)of 233.09 addresses per second.Compared with existing methodologies,6EDL-N exhibits a remarkable improvement in hit rate,which is 8.13 times higher than current methods,and an NPT that is 14.94 times higher than the current methods.Additionally,its coverage is 1.84 times greater than that of existing methods.In the scenario with seed addresses,we propose the 6EDL-S method for active IPv6 address detection based on adversarial networks.This method leverages precise learning of seed address distribution patterns and employs an environmental feedback mechanism to mitigate seed address sampling biases,effectively enhancing the hit rate.Experimental results demonstrate that the hit rate of 6EDL-S reaches 25.91%,which is 1.23~10.89 times higher compared with existing methods.Furthermore,NPT is 466.72 addresses per second,which is 1.49~6.20 times higher than existing methods.Ultimately,through continuous active IPv6address probing,the 6EDL system successfully discovered 29.77 billion active IPv6 addresses,comprising 5.66 billion alias addresses and 24.11 billion non-alias addresses,encompassing 125101 BGP prefixes and 40137 autonomous systems.The active IPv6 address set(IPv6 hitlist)constructed in our work will effectively support various applications such as IPv6 network measurement and security analysis,further opening the door to IPv6 network research.
作者
宋光磊
张文健
林金磊
韩东岐
王之梁
张辉
杨家海
SONG Guang-Lei;ZHANG Wen-Jian;LIN Jin-Lei;HAN Dong-Qi;WANG Zhi-Liang;ZHANG Hui;YANG Jia-Hai(The Institute for Network Sciences and Cyberspace,Tsinghua University,Beijing 100084;Zhongguancun Laboratory,Beijing 100084;Quancheng Laboratory,Jinan 250000)
出处
《计算机学报》
EI
CAS
CSCD
北大核心
2024年第8期1949-1969,共21页
Chinese Journal of Computers
基金
国家重点研发项目互联网IP地址空间与域间路由系统关键信息感知技术(No.2022YFB3105001)
中关村实验室项目、清华大学-中国电信集团有限公司下一代互联网技术联合研究中心项目资助.
