A principled approach to measuring the IoT ecosystem

Kumar, Deepak

Permalink

https://hdl.handle.net/2142/108499 Copy

Description

Title

A principled approach to measuring the IoT ecosystem

Author(s)

Kumar, Deepak

Issue Date

2020-07-15

Director of Research (if dissertation) or Advisor (if thesis)

Bailey, Michael

Doctoral Committee Chair(s)

Bailey, Michael

Committee Member(s)

• Borisov, Nikita
• Bates, Adam
• Wang, Gang
• Durumeric, Zakir

Department of Study

Computer Science

Discipline

Computer Science

Degree Granting Institution

University of Illinois at Urbana-Champaign

Degree Name

Ph.D.

Degree Level

Dissertation

Keyword(s)

• internet measurement
• security
• IoT
• networking
• computing

Abstract

Internet of Things (IoT) devices combine network connectivity, cheap hardware, and actuation to provide new ways to interface with the world. In spite of this growth, little work has been done to measure the network properties of IoT devices. Such measurements can help to inform systems designers and security researchers of IoT networking behavior in practice to guide future research. Unfortunately, properly measuring the IoT ecosystem is not trivial. Devices may have different capabilities and behaviors, which require both active measurements and passive observation to quantify. Furthermore, the IoT devices that are connected to the public Internet may vary from those connected inside home networks, requiring both an external and internal vantage point to draw measurements from. In this thesis, we demonstrate how IoT measurements drawn from a single vantage point or mesaurement technique lead to a biased view of the network services in the IoT ecosystem. To do this, we conduct several real-world IoT measurements, drawn from both inside and outside home networks using active and passive monitoring. First, we leverage active scanning and passive observation in understanding the Mirai botnet---chiefly, we report on the devices it infected, the command and control infrastructure behind the botnet, and how the malware evolved over time. We then conduct active measurements from inside 16M home networks spanning 83M devices from 11~geographic regions to survey the IoT devices installed around the world. We demonstrate how these measurements can uncover the device types that are most at risk and the vendors who manufacture the weakest devices. We compare our measurements with passive external observation by detecting compromised scanning behavior from smart homes. We find that while passive external observation can drive insight about compromised networks, it offers little by way of concrete device attribution. We next compare our results from active external scanning with active internal scanning and show how relying solely on external scanning for IoT measurements under-reports security important IoT protocols, potentially skewing the services investigated by the security community. Finally, we conduct passive measurements of 275~smart home networks to investigate IoT behavior. We find that IoT device behavior varies by type and devices regularly communicate over a myriad of bespoke ports, in many cases to speak standard protocols (e.g., HTTP). Finally, we observe that devices regularly offer active services (e.g., Telnet, rpcbind) that are rarely, if ever, used in actual communication, demonstrating the need for both active and passive measurements to properly compare device capabilities and behaviors. Our results highlight the need for a confluence of measurement perspectives to comprehensively understand IoT ecosystem. We conclude with recommendations for future measurements of IoT devices as well as directions for the systems and security community informed by our work.

Graduation Semester

2020-08

Type of Resource

Thesis

Permalink

http://hdl.handle.net/2142/108499

Copyright and License Information

Copyright 2020 Deepak Kumar

Owning Collections