Automatic Transmission – a data-privacy study of connected vehicles

Your car is a smartphone on wheels. Here's who's listening.

The first large-scale measurement study of the connected-vehicle ecosystem.

A connected car is a vehicle with built-in internet access — Wi-Fi, cellular, GPS — that lets it communicate constantly with its manufacturer and outside companies. Over 75% of vehicles sold globally have this connectity built-in.

Photo is not loadingimages/connected-car.jpg

It knows where you drive.

It knows who you are.

It can share this data and more with insurance companies, advertisers, etc...

Research Questions

Photo is not loading images/car-diagram.png

Connected Vehicle Ecosystem

This diagram shows the data flows to and from a vehicle and its companion mobile app. Both devices send data, including private consumer data, to different 1st and 3rd party servers using Wi-Fi and cellular service. Solid arrows represent flows that were intercepted through our experiments.

The Problem

Once the data gets sent to these servers, it is up to the companies that receive the consumer information to make decisions on what they do with it. Unfortunately, in many cases, this includes sharing or selling consumer data to other undisclosed 3rd parties.
Consumers have no control over their data once it has left their device

In this paper, we take the first steps to address the limited visibility into the privacy implications of the connected vehicle ecosystem. We identify two vantage points in the ecosystem where we can gain insight into the data that connected vehicles are sharing with both manufacturers and third parties: the vehicles themselves and the mobile apps provided by manufacturers.
We ask the following guiding questions:

• 01What personal consumer data do connected vehicles and their companion mobile apps transmit?
• 02Who receives that personal consumer data?
• 03What is the manufacturer response to these findings?

Methods

We investigated 21 vehicles from the U.S. market in a controlled environment along with 30 companion mobile apps instrumented with on-site vehicles between October 2024 and August 2025. Below is a description of the experiments we ran and our setup.

Vehicle Testing

Photo is not loading images/rpi.png

Wi-Fi Testing Setup

To collect Wi-Fi traffic from vehicles, we configured a custom access point (AP) on a Raspberry Pi and used tcpdump to log all packets that were sent or received via this AP.
This allowed us to see all the destinations the vehicles were sending data to but not the information within the packets as it was encrypted.

Photo is not loading images/test-facility.jpg

Stationary Tests

Idle Baseline
vehicle on — no activity

Active Test
perform all possible actions

Driving Test

drive 5–45 mph with acceleration & hard braking

Isolating Cellular Traffic

Photo is not loading images/faraday-tent.jpg

One hypothesis we tested was whether blocking a vehicle’s ability to communicate over its cellular network would force more Wi-Fi communication. To block external cellular signals, we drove 11 EVs in the sample into a car-sized Faraday tent providing ≈93 dB of attenuation, blocking their cellular connection entirely. Stationary Idle and Active tests were repeated inside the tent to see whether traffic that normally goes out over cellular rerouted to Wi-Fi instead.

App Testing

In total, we experimented with 30 connected vehicle companion apps that were paired with the vehicles at Consumer Reports’s testing facility.

Device Setup

• 01We used a combination of test phones and iOS versions for our experiments: an iPhone 8/iOS 16.6, an iPhone X/iOS 16.7.11, and an iPhone 13/iOS 18.5.
• 02To minimize background traffic, we deleted all non-essential apps on the test phones and tested apps one-by-one, including deleting each vehicle app and restarting the phone before downloading the next app.
• 03We used the iOS native screen recording feature to record our interactions with each app for later review.
• 04To capture and decrypt network traffic from the companion mobile apps, we used iPhones with custom root certificates connected to mitmproxy.

Process for Testing Each App

• 01During app installation and login we accepted all permission requests (e.g., tracking, location, calendar access, Bluetooth, notifications) that the application requested
• 02We had a Consumer Reports employee log into the app using their existing credentials associated with a vehicle on the lot.
• 03Once we were logged-in, we manually exercised all available functionality, such as looking for nearby charging staions, geolocating the vehicle, viewing vehicle data and service history (e.g., tire pressure), viewing notifications (e.g., “doors are unlocked”), and viewing in-app privacy policies.
• 04Some apps allowed us to perform physical interactions on the vehicle, such as remotely opening the trunk. We performed all such actions and verified that the vehicle completed each request.

Vehicle Dataset

Findings

• 19 of 21 vehicles contacted at least one third party over Wi-Fi, including known advertising and tracking domains
• 7 of 30 companion apps transmitted sensitive identifiers (VINs, emails, phone numbers, precise location) to third parties associated with advertising and tracking.
  Transmiting multiple forms of PII to the same third party allows advertisers to build in-depth profiles on consumers

• Pairing a companion app roughly doubled a vehicle's exposure to advertising/tracking companies on average, and in some cases added 20+ new ones.

Manufacturer Disclosures

• The team disclosed these findings to the different manufacturers featured in the study, below is an interactve summary of the different explanations received.

• The ongoing theme of all these responses was shifting the blame to the consumer.
• The current system does not give owners the ability to choose.
  Assuming owners can find the particular agreements, if an owner decides they are uncomfortable with the data sharing described within them, they are faced with (arguably) unfair choices:
  01accept the agreements regardless of their concerns
  02stop using their car's connected-vehicle features - which includes remote start, the app, and other very useful features
  03stop using the vehicle entirely
  
• As one notable exception, Honda improved its data collection practices to prevent sending precise geolocation to a third party associated with user tracking.

Conclusion

• We found that vehicles, under a variety of real-world settings, contact not only a wide range of first parties (i.e., manufacturer domains) and car-specific support parties, but also third parties that are known to provide advertising and tracking services.
• Vehicles from the same manufacturer exhibit different network behaviors - this makes it very difficult and expensive to study these vehicles.
• When adding vehicle companion apps to the analysis, we found that vehicle owners are exposed to even more privacy-sensitive ATA communication—in some cases more than two dozen additional trackers.
• Our study revealed a large gap between what vehicle manufacturers publicly disclosed and how the connected vehicle ecosystem actually shares data over the Internet
• Based on the opaque nature of vehicular systems, we argue that there is a need for better transparency to ensure increased visibility into the entire ecosystem to identify and address corresponding harms.