The Tesla Model Y is the best-selling vehicle in the world in the first quarter of 2023
©Lillian SUWANRUMPHA / AFP
Depending on which newer model vehicle you own, your car may be watching you – literally and figuratively – while you drive.
It watches you thanks to cameras that monitor the cabin and detect where you are looking, or thanks to sensors that track your speed, your position in the lane and your rate of acceleration.
Your car uses this data to make your driving safe, comfortable and convenient. For example, cameras can determine if you were distracted and need to refocus your attention on the road. They can also detect speeding by checking the speed limit based on your GPS location or traffic signs along the road and warn you to slow down. Some automakers are also starting to incorporate similar convenience features, such as unlocking your car by scanning your face or fingerprint. Your car may also transmit some of this data to the manufacturer’s data centers, which will use it to improve your driving experience or provide you with personalized services.
Aside from these benefits, this data collection is a privacy nightmare. This information may reveal your identity, your driving habits, driving safety, the places you have been and where you regularly go. A report from the Mozilla Foundation, a nonprofit technology research and advocacy organization, concluded that automakers’ privacy policies are extremely lax. The study identifies cars as the “worst privacy product category we have ever studied.” On November 30, 2023, US Senator Ed Markey wrote a letter to US automakers asking a long series of questions about their data practices.
Today’s smart cars present drivers with a trade-off between convenience and privacy, assuming drivers have the ability to improve their cars’ privacy. As a computer scientist working on cybersecurity and transportation resilience, I see several technological paths to get the best of both worlds: cars that use this collected data while maintaining user privacy.
Today’s cars use a variety of sensors to understand the environment, analyze data and ensure passenger safety. For example, cars are equipped with sensors that measure brake pedal position, vehicle speed, driver movements, surrounding vehicles and even traffic lights. The collected data is then transmitted to the car’s electrical control units.
There are two types of sensors that continuously monitor and predict driver fatigue. The first consists of sensors for monitoring vehicle health, such as lane detection and steering wheel position tracking. This data is not directly linked to a specific individual and may be considered personal data unless it is linked to other data that identifies the driver.
The second type of sensors makes it possible to track the drivers themselves. This category includes, for example, cameras that track the driver’s eye movements to predict fatigue. This second group of sensors is directly related to driver privacy as it collects personally identifiable information such as the driver’s face.
Protection of private life
Depending on the scope of services and features, there is a trade-off between the quality of the driving experience and the protection of the driver’s privacy. Some drivers will prefer to share their biometric data to make it easier to access vehicle features and automate much of their driving experience. Others will prefer to manually control the car’s systems and share less or no personal data.
At first glance, it seems that the compromise between privacy and driver comfort cannot be avoided. Automakers tend to take measures to protect driver data from data thieves, but they collect a lot of data themselves. And as the Mozilla Foundation report shows, most automakers reserve the right to sell your data. Researchers are working to develop data analysis tools that better protect privacy and make progress toward eliminating the trade-off.
For example, over the past seven years, the concept of federated machine learning has gained attention because it allows algorithms to learn from data on your local device without copying the data to a central server. For example, Google’s Gboard keyboard benefits from federated learning to better guess the next word you’re likely to type without sharing your private data with a server.
The study, led by Ervin Moore, a graduate student in Florida International University’s Sustainability, Optimization, and Learning for InterDependent Networks Lab, and published in the IEEE Internet of Things Journal, explored the idea of using a federated blockchain-based machine learning platform to Improve privacy and security of users and their sensitive data. This technology could be used to protect driver data. There are also other privacy protection techniques such as: B. Location obfuscation, which changes the user’s location data to prevent their location from being revealed.
Although there is always a trade-off between privacy and quality of service, privacy-preserving data analysis techniques could pave the way for the use of data without exposing the personal information of drivers and passengers. This would allow drivers to benefit from a wide range of services and functions of modern cars without having to pay the high price of losing their privacy.
The original version of this article was published on The Conversation.