AI Self-Driving Cars And Blurring Of Sensory Data For Privacy
About this episode
AI self-driving cars rely on constant sensor capture, creating a “roving eye” that could reconstruct where people go and when—far beyond today’s smartphone or roadside cameras. Dr. Lance Eliot explores whether intentional data blurring can make this surveillance more acceptable without harming driving performance, keeping the foreground clear while obscuring background details. He weighs on-car vs cloud blurring, encryption and upload policies, and the risk that blurring could be reversed via evolving “blur/unblur” cat-and-mouse algorithms.
Dr. Eliot explores how the sensory data from AI self-driving cars can be intentionally blurred to help protect the privacy of those seen and heard by the roaming AI. See his Forbes column for further info: https://www.forbes.com/sites/lanceeliot/
data blurring
"In this episode, I'll be discussing the topic of using data blurring for self-driving cars. ... The use of a blurring technique can potentially be used as a means of trying to attain a level of privacy for those that might be captured on an image or video or altogether in data."
Data blurring is when a system intentionally makes parts of an image or video harder to recognize. For self-driving cars, that can help hide faces or other identifying details. The car still tries to understand the road, but with less personal information exposed.
Data blurring is a privacy technique where identifying details in sensor data (like video or images) are intentionally obscured. The goal is to reduce how well a system can recognize people or sensitive information while still allowing useful driving-related perception. In self-driving contexts, it’s meant to protect bystanders captured by vehicle cameras and other sensors.
self-driving cars
"Let's talk about cars. The future of cars consists of self-driving cars. These are cars that have an AI driving system at the wheel of the vehicle."
Self-driving cars are cars that use computers to do the driving. Instead of a person controlling everything, the car’s AI handles things like speed and steering. To do that, it uses sensors to “see” what’s around it.
Self-driving cars are vehicles that use an AI driving system to perform driving tasks rather than relying on the human driver for steering, acceleration, and braking. They depend on sensors to understand the environment and make driving decisions. This episode frames them as continuously capturing imagery and other data as they operate.
AI driving system
"These are cars that have an AI driving system at the wheel of the vehicle. The driving actions are undertaken by AI."
An AI driving system is the car’s computer brain. It looks at what the sensors detect and decides what the car should do next. In this context, it’s what actually performs the driving actions.
An AI driving system is the onboard computer software that interprets sensor inputs and decides what driving actions to take. In self-driving cars, it replaces the driver’s real-time control by generating steering, speed, and braking commands. The episode emphasizes that the driving actions are undertaken by AI.
various sensors
"An important element of self-driving cars is the use of various sensors to detect the driving scene. Though this seems relatively innocent and there isn't much attention being given to the plethora of sensors that self-driving cars contain."
Self-driving cars use multiple sensors—like cameras or other detectors—to understand what’s happening around the vehicle. The episode’s point is that these sensors can collect a lot of detail. That’s why privacy becomes an issue.
Self-driving cars use various sensors to detect and interpret the driving scene. These sensors can capture far more information than people expect, which is central to the privacy concern raised in the episode. The speaker notes that there isn’t much attention paid to the “plethora of sensors” these cars contain.
roving eye
"I've referred to this as the roving eye of the coming era of self-driving cars. Do you want just anyone to know where you were last Monday or Tuesday?"
“Roving eye” means the idea that self-driving cars could constantly watch the world as they travel. That raises privacy worries because it could reveal where people go.
“Roving eye” is a metaphor for how self-driving cars could continuously observe and collect information across many locations. The concern is that widespread sensing could enable reconstruction of people’s movements and routines at scale.
sensory data
"In a sense, that's peanuts in comparison to the magnitude of video and other sensory data captured by self-driving cars."
Sensory data is the information the car gathers from its “eyes and ears,” like cameras and other sensors. The car uses it to figure out what’s happening around it.
Sensory data is the raw information collected by a vehicle’s perception system—typically from cameras, radar, lidar, and microphones. For self-driving cars, this data is the foundation for detecting objects, estimating their motion, and understanding the driving scene.
video cameras
"Some people are already worried about privacy intrusion from video cameras that are mounted on telephone poles, or that are used by people as they carry around their smartphones."
Video cameras are imaging sensors that capture visual information (frames) used for perception and recording. In the privacy context, they can reveal identities, locations, and activities when their footage is stored, transmitted, or analyzed.
blurring techniques
"Well, the earlier discussion about the blurring of information or data was, in fact, the answer to this potential problem. Here's an intriguing question to ponder. Will the advent of AI-based, true self-driving cars and the roving eye be potentially made more societally palatable via the use of blurring techniques?"
Blurring techniques are ways to hide identifying details in images, like faces or license plates. The goal is to protect privacy without ruining the car’s ability to understand the road.
Blurring techniques are privacy-preserving image-processing methods that obscure identifiable details in captured footage or sensor outputs. The idea here is to reduce personal privacy risk while still keeping enough scene detail for safe driving perception.
foreground
"There's not much debate that the foreground of any sensory detection by a self-driving car is likely to be crucial for the driving of the vehicle. As such, the foreground is going to have to be kept in focus."
The foreground is the area in front of the car that matters most for driving. If the car can clearly see the foreground, it can react to nearby hazards better.
In vehicle perception, the foreground is the part of the scene that is closest and most immediately relevant to driving decisions. Keeping the foreground “in focus” helps the system accurately detect and track nearby objects like vehicles, pedestrians, and animals.
background
"The more open-ended question is whether the background needs to be kept in focus too."
The background is everything farther away behind the main objects the car is reacting to. Sometimes it needs to be clear for navigation and context, but sometimes it can be treated differently.
In perception systems, the background is the farther portion of the scene behind the most immediate objects. Whether it must remain sharp depends on the task—e.g., recognizing signs, lane context, or tracking objects that move into the foreground.
intentional blurring
"This is where intentional blurring comes to play. One approach consists of taking the data after it has been examined for driving purposes, and then blurring the data so that those elements are generally considered irrelevant to the driving act can no longer be discerned."
Intentional blurring means the system intentionally “messes up” parts of the sensor information to protect privacy. It tries to keep the useful driving details, while hiding things that could identify people.
Intentional blurring is a privacy technique where a self-driving system modifies sensor-derived data so that identifiable or non-driving-relevant details can’t be recognized. The goal is to reduce what could be used to track people or infer sensitive information while still keeping enough information for safe driving.
onboard system
"If you do so while the data is fresh and just brought into the onboard system, this means that you need the computational resources onboard the car to do this type of blurring action."
The onboard system is the car’s in-vehicle computing hardware and software that receives sensor inputs and performs perception and driving-related processing. The transcript highlights that blurring “while the data is fresh” requires enough onboard compute power to do it in real time.
cloud
"Another notion is you could do the blurring once the data has been uploaded into the cloud. This brings up there are two camps at this."
Here, “cloud” means data is sent to remote computers over the internet. The debate is whether the car should blur data before sending it, or send it first and blur it after it reaches those remote servers.
In this privacy discussion, “cloud” refers to remote servers where sensor data might be uploaded for processing, storage, or analysis. The transcript contrasts doing blurring before upload versus uploading first and blurring later, depending on different privacy philosophies.
unblurred data
"One camp says the full and unblurred data should never be allowed to leave the car. In that sense, it should not be allowed to be uploaded to the cloud."
Unblurred data is the original, not-yet-privacy-protected information from the sensors. The debate is whether that raw data should stay in the car or be sent somewhere else after additional safeguards.
“Unblurred data” is sensor data that has not been privacy-modified, meaning identifiable or sensitive elements may still be present. The transcript contrasts two approaches: keeping full unblurred data inside the car versus allowing it to be uploaded after encryption and/or later blurring.
encryption
"Only once it has been blurred and possibly encrypted could it be uploaded. The other camp says it's find up one of the whole shebang. As long as it's encrypted, you just blur it after it gets into the cloud."
Encryption is like putting data into a locked code. Even if someone gets the data, they can’t understand it without the proper key.
Encryption is the process of transforming data into a coded form so it can’t be read without the right key. In the transcript, encryption is paired with blurring to reduce the risk of sensitive sensor data being accessed if it’s uploaded.
blurring algorithms
"Sometimes that which can be blurred can later on be unblurred. This means that the blurring might be undone, mathematically or computationally. There is an ongoing cat and mouse gambit of blurring algorithms that go to battle with each other."
Blurring algorithms are the rules or computer methods used to hide sensitive details in the data. The tricky part is that other methods can sometimes undo that hiding, so the “hide” and “undo” tools keep improving.
Blurring algorithms are the mathematical methods used to obscure parts of sensor-derived data so that sensitive details can’t be identified. The transcript emphasizes that these methods can sometimes be reversed, which drives an ongoing improvement cycle between blurring and unblurring techniques.
Request an Explanation
Heard something you'd like explained? We'll add it to this episode.
Sign in to request explanations for terms you heard.
Want to learn more?
Browse our glossary for plain-English explanations of automotive terms, jargon, and concepts.
Help improve this episode
See something that's not quite right? Our annotations are AI-generated and can sometimes miss the mark. Click the flag icon on any annotation to suggest a correction.