Fast Internet-Wide Scanning: A New Security Perspective

[摘要] Techniques like passive observation and random sampling let researchers understand many aspects of Internet day-to-day operation, yet these methodologies often focus on popular services or a small demographic of users, rather than providing a comprehensive view of the devices and services that constitute the Internet. As the diversity of devices and the role they play in critical infrastructure increases, so does understanding the dynamics of and securing these hosts. This dissertation shows how fast Internet-wide scanning provides a near-global perspective of edge hosts that enables researchers to uncover security weaknesses that only emerge at scale.First, I show that it is possible to efficiently scan the IPv4 address space. ZMap: a network scanner specifically architected for large-scale research studies can survey the entire IPv4 address space from a single machine in under an hour at 97% of the theoretical maximum speed of gigabit Ethernet with an estimated 98% coverage of publicly available hosts. Building on ZMap, I introduce Censys, a public service that maintains up-to-date and legacy snapshots of the hosts and services running across the public IPv4 address space. Censys enables researchers to efficiently ask a range of security questions.Next, I present four case studies that highlight how Internet-wide scanning can identify new classes of weaknesses that only emerge at scale, uncover unexpected attacks, shed light on previously opaque distributed systems on the Internet, and understand the impact of consequential vulnerabilities. Finally, I explore how in- creased contention over IPv4 addresses introduces new challenges for performing large-scale empirical studies. I conclude with suggested directions that the re- search community needs to consider to retain the degree of visibility that Internet-wide scanning currently provides.

[发布机构] University of Michigan

[效力级别] Computer Science

[关键词] Internet-Wide Scanning;Computer Science;Engineering;Computer Science & Engineering