Where Things Roam: Uncovering Cellular IoT/M2M Connectivity

Andra Lutu, Byungjin Jun, Alessandro Finamore, Fabián E. Bustamante, Diego Perino

Abstract

Cellular IoT devices, or "things," are increasingly roaming globally, often permanently. This paper presents the first large-scale characterization of the global cellular IoT roaming ecosystem. By analyzing massive datasets from a global M2M provider and a major European operator, the authors uncover that IoT devices act as "silent roamers"—stationary devices that remain active on visited networks far longer than human roamers and generate disproportionately high signaling overhead. The study highlights critical challenges for mobile network operators in managing these stationary but technically roaming devices.

Introduction

While international roaming is traditionally associated with human travel, a growing portion of cellular traffic originates from Machine-to-Machine (M2M) devices like connected cars, smart meters, and industrial sensors. These devices are often manufactured in one country but deployed globally, relying on "IoT SIMs" provided by M2M platforms that allow them to roam on hundreds of networks indefinitely. This paper provides a rare, large-scale look into this "opaque" world of international IoT roaming.

The Roaming Challenge

Manufacturers utilize M2M platforms to simplify global connectivity through a single SIM. This creates a unique burden for Visited Mobile Network Operators (VMNOs)—the networks these devices actually connect to in the field. Unlike human tourists who typically leave after a short period, IoT devices often "permanently roam," staying on a visited network for years.

Overview of cellular roaming architectures, illustrating the paths for home-routed, local breakout, and IPX breakout traffic.

As shown in the architectural overview, roaming traffic can be routed back to the Home MNO (HMNO) or broken out locally. Because these devices exhibit traffic patterns—such as high signaling relative to low data volume—they can stress network infrastructure in ways the system wasn't originally designed to handle. Identifying these "things" among millions of devices remains a significant hurdle for operators attempting to manage resources or negotiate commercial terms.

Methodology

The researchers developed a multi-step classification approach to identify and categorize IoT devices within massive datasets:

  • TAC Analysis: Identifying device types via the Type Allocation Code (the first 8 digits of an IMEI).
  • APN Keyword Mapping: Searching for M2M-specific Access Point Names (e.g., "scania" for automotive devices).
  • Traffic Profiling: Distinguishing smartphones from IoT devices based on signaling-to-data ratios, mobility (radius of gyration), and session duration.
  • Roaming Labeling: Tagging devices based on their Home MNO (HMNO) versus the network they are currently using to track roaming status.

The study leveraged two primary datasets: one from a global M2M provider covering 120,000 4G devices across 70 countries, and another from a major UK operator capturing 39.6 million unique devices, of which 10.1 million were identified as M2M-related.

Key Insights

Extreme Centralization

The global IoT roaming ecosystem is surprisingly concentrated. A few "home" countries provide the vast majority of roaming devices. For the M2M platform studied, just three countries (the Netherlands, Sweden, and Spain) provided the "home" for 60% of all inbound roaming IoT devices.

Heatmap showing the global distribution and concentration of home countries (HMNO) for roaming devices across various visited countries.

The Stationary Roamer Paradox

Despite being technically classified as "roaming," the study found that 80% of these devices are stationary, with a radius of gyration of less than 1km. They are permanent fixtures of the visited network rather than transient visitors. Furthermore, the median M2M device stays active on a visited network 4.5x longer than a roaming smartphone.

Signaling Stress and the 2G Legacy

Roaming IoT devices generate approximately 10x more signaling messages than native devices. This "background noise" consumes network resources without necessarily generating data revenue for the operator. Additionally, despite the push toward 5G, a staggering 77.4% of M2M devices in the MNO dataset were active only on 2G networks, suggesting that sunsetting legacy infrastructure will be significantly delayed by the long lifecycle of IoT hardware.

Conclusion

IoT roaming presents a paradigm shift for mobile operators. The "silent roamer" phenomenon requires new management strategies and pricing models that account for signaling load and long-term network residency rather than just data volume. Future research is needed to explore the impact of 5G and specialized protocols like NB-IoT as they become more widely deployed.

Summary updated

BibTeX
@inproceedings{lutu2020where,
  author = {Andra Lutu and Byungjin Jun and Alessandro Finamore and Fabián E. Bustamante and Diego Perino},
  title = {{Where Things Roam: Uncovering Cellular IoT/M2M Connectivity}},
  booktitle = {Internet Measurement Conference},
  year = {2020},
  doi = {10.1145/3419394.3423661},
}
Figure