Dutch forensic lab says it has decoded Tesla's driving data
reuters.com
reuters.com
To combat fundamental attribution error, whenever you read an article about Tesla, try replacing the word 'Tesla' with your least favorite car or consumer product company.
'Brand X' has a bunch of data about accidents, and they are motivated to not share all of it with authorities. When they do share data they are motivated to share the data that presents them in the best light. How would it feel if there was data about a crash you were involved in that you could not get access to?
First extension I see: do this with people :D
E.g. in China they don’t trust parents with data on the gender of their child from ultrasounds due to the prevalence of sex-selective abortions. In other countries there are parties about the data. Data is always in a cultural context.
I'd be angry that data was recorded in the first place. No data should be recorded without the driver's explicit consent.
Why do you think they had all those sensors in the early models without the software to use it?
In Teslas, there are two main CAN busses: the regular one and the 'secure' one that does the brakes and other safety stuff. There is NO connection between the two; but there is a secured path where the gateway can listen & get data from the secure one. The point here is that vehicle safety CAN bus is likely unhackable remotely, as there simply is no RF path writing onto that bus. (all bets are off if you have the vehicle, physically). I believe all previous exploits have been fixed.
And one more thing: ECUs have some amount of on-board flash that is used during a crash/power loss scenario. The "dump important stuff to flash" CAN command causes the ECU to take an NMI-style interrupt and do it. Also, if the ECU's own power monitor notices the ECU is about to lose power, there is enough energy storage on board to operate for maybe a few hundred ms. The important stuff is saved.
Every ECU must have a way to OTA its code. E V E R Y O N E, including some door lock or whatever (essentially LIN to CAN gatewaying). There are no, as in ZERO, exceptions to this rule. (or, there weren't as of five years ago, ymmv)
Except if there is some kind of encryption, I suppose.
The issue was that Tesla was only providing certain data that was being asked for.
You wouldn't know that it would be possible to ask for more information, unless you knew they were collecting it.
Now they know Tesla has this information they should be able to get it from them regardless if they have to decrypt it themselves or not.
(Unless Tesla decides to stop collecting data for some reason.)
Only if Tesla purjured themselves during discovery. That's not the way court cases work. You can ask "Do you guys record the accelerometer data anywhere else but on the MCU?" and they have to answer truthfully.
What you write here mostly amounts to a conspiracy theory. You're taking the idea that Tesla is "hiding something nefarious" as a prior and then taking evidence that their storage was reverse engineered as evidence for that prior.
But that's not how engineering works either.
But I suppose Tesla isn't hiding anything nefarious. Most of their malfeasance is plain to see.
Most legal precedent I’ve seen says a lot less time than they’ve already delayed it, frankly.
Everyone knows that Elon Musk's twitter is not a reliable indicator of feature release dates.
First part yes, second part no. It has been in development since at least 2015 and still drives straight into walls. Meanwhile Waymo and Cruze are putting down tens of thousands of miles between interventions.
It's fraud when someone is harmed. People like you who didn't buy the product and don't want it don't have standing here. People like me, who did, do. And almost without exception, we're all lined up to get the beta release. Offer me a $10k refund today in exchange for never getting access to FSD releases, and I'd turn it down.
"People willfully paying for something I don't like" isn't fraud. Find a FSD owner willing to sue for a refund.
It wouldn't. It would be an existence proof that this isn't fraud: I want the product, they offer it as a pre-purchase with disclaimers, I bought it, ergo I wasn't harmed. Getting a refund now for a product a fully intend to purchase again in the future is called a "loan".
Again, fraud is a crime. It requires damages. I have to have been harmed.
Yes, it would. Because then I would want a refund. But that would be asserting facts not in evidence, and I refer to my point upthread about arguing from a conspiracy as a prior.
Being convicted of drug possession being one really clear example - there is often no need to even prove the person was aware of the existence of the drugs. Mere actual possession is usually enough (depending on the jurisdiction). And there is no need to prove the person intended to sell, distribute, harm, or even do anything with them. Even if the person literally was taking them to be destroyed, they can still be convicted of the crime.
(a) Except as otherwise provided in this section, whoever, in any matter within the jurisdiction of the executive, legislative, or judicial branch of the Government of the United States, knowingly and willfully—
(1) falsifies, conceals, or covers up by any trick, scheme, or device a material fact;
(2) makes any materially false, fictitious, or fraudulent statement or representation; or
(3) makes or uses any false writing or document knowing the same to contain any materially false, fictitious, or fraudulent statement or entry;
So if Tesla is knowingly taking pre-orders for FSD without being certain on when they could deliver it _and_ they don't make it clear that they don't know when or if FSD could be delivered, that would likely qualify as fraud.
And given the length of time it’s been going, at some point someone is going to call time and it’s going to be a big mess for Tesla. Unless they actually deliver it.
No pressure!
2. Tesla is currently advertising their cars as being capable of driving themselves. They have videos of people driving with their hands off the wheel. They have failed to implement safety features that make drivers pay attention while using driver assists. People have died because of this.
3. There are no FSD owners because FSD does not exists.
4. You are a victim too. You will realize it someday.
5. I sure as hell have standing. Have you seen the video of the current FSD version? It just plain does not work and is being released on public roads for "testing" by untrained drivers. Tesla is putting my life in danger, and shame on you for going along with it.
If they know it’s there and can proactively argue it’s appropriate, that changes the equation dramatically.
Fundamentally this process isn't any different than reverse engineering. If a good hacker can figure out how something works based on knowledge of the problem domain and technology in use, then those same questions can be answered in a deposition and provided via discovery, by definition.
AGAIN you are arguing from a conspiracy as a prior. There is no "information" to "get out". The car stores telemetry and sensor input. We always knew that. This is a story about someone figuring out how to recover it.
Rather no one dumps everything in a deposition. If there is something that could help opposing counsel, then it will be avoided as best as possible. This is basic law 101.
Just because you, and some others, know exactly what is there (presumably) doesn’t mean everyone else does! And I’ve seen first hand how opposing counsels can screw themselves over by not spending time and effort to figure out how the technology actually works, and many times plaintiffs don’t either, so they just don’t know where to poke.
And that is avoiding that knowledge of a previously hidden area (such as if something is recoverable that seems like it should not be because they have the wrong assumptions or don’t know how it actually works internally!) can radically change the actual strategy and applicable law.
You seem to keep asserting that everyone already knows everything and all optimal strategies are already known because of this, which does not apply to the real world at all.
Totally. I've played this game with FOIA (public records requests). It's exhausting. Of course they had the data. But as a private citizen, I had to prove they had the data before I could request it.
Tesla really seems to have an extreme amount of personal data capture combined with always-on remote access.
It’s fascinating to watch people swear off Amazon Echo and delete their Facebook over privacy concerns but then proudly drive around in a vehicle that records everything, stores everything, and gives opaque data access to the parent company. In this case, a parent company with a reputation for bending the rules to suit their needs.
Really goes to show how much the computing privacy debate is a matter of brand loyalties and each company’s position on the hype cycle more than anything else.
GDPR, while not perfect, describes "personal data" to be:
> The data subjects are identifiable if they can be directly or indirectly identified, especially by reference to an identifier such as a name, an identification number, location data, an online identifier or one of several special characteristics, which expresses the physical, physiological, genetic, mental, commercial, cultural or social identity of these natural persons.
> Personal data are any information which are related to an identified or identifiable natural person.
Tesla already knows who you are so any data they capture is personal data.
That's not how that works. Just because IP is personal data doesn't mean that the event "user clicked on this button" is personal data...
https://www.mpgomatic.com/2007/11/04/toyota-corolla-gas-mile...
Reasons for this are probably creeping sizes and weights and larger engines, larger tires and softer rides.
There is a crossover point after you drive a certain number of miles in an EV, after which the lifetime carbon cost is lower than that of an equivalent ICE car.
For Model 3 the crossover point is estimated at around 5,000 miles. Not "many years."
The sources and methodology for this estimate are discussed in detail in Tesla's 2020 Impact Report.
https://www.tesla.com/ns_videos/2020-tesla-impact-report.pdf
Excerpt:
—
Variables often overlooked by other lifecycle studies:
* Using Worldwide Harmonized Light Vehicle Test Procedure (WLTP) or Environmental Protection Agency (EPA) fuel/energy consumption data (both of which overestimate fuel-economy and underestimate emissions) rather than real-world data;
* Not considering the higher energy efficiency of Tesla’s powertrains;
* Assuming the average EV needs a battery replacement at some point in its life;
* Not considering emissions generated through the oil refining and the transportation process; and
* Using outdated data for the carbon impact of cell manufacturing.
I think 5k miles is a bit low, and using tesla as your only source biased.
Putting that up against Tesla, which would have a lot to lose by putting the wrong analysis into their Impact Report, I would give Tesla more credibility despite the fact that they have a vested interest here.
Their vested interest is out in the open, not a secret… it's the mission of the company even.
Regardless, the discussion wasn't about replacing an ICE car with an electric, it was about replacing an old ICE car with a newer ICE one.
I will also note that personally I have a a very low level of trust for any statements put out by Musk's companies - they have a clear history of greatly exaggerating various good news.
Perhaps chassis costs (meaning carbon costs of making the chassis) are roughly comparable between ICE and EVs, so I suppose there was no need to focus on this in the report.
>and engine parts, it seems that those are acquired
Tesla motors are wound and made from scratch from raw materials on site. This data point would not be in favor of ICE cars, for the reasons you mention. Batteries on the other hand do increase the carbon cost and that's why EVs start off behind ICE cars at the start of their lifetime, catching up in terms of lower carbon used only after some period of use.
>Finally, the 24.8MPG number they cite is pretty large
They go by real world consumption, not manufacturer claims. Remember VW Dieselgate? Credibility is not a given in ICE circles. But you have a good point it's probably more about the EU/US difference here with US cars being gas hogs, due to size and possibly the impact of different emissions equipment in some states.
>Regardless, the discussion wasn't about replacing an ICE car with an electric
OK. It was one line, not so much a discussion, but it did come from you to be fair, so although I didn't read it that way, you get to decide. Still, I believe the comparison makes most sense against an EV, although it may not have been what you had in mind. If one cares enough about carbon to even make the comparison, and to try to defend the position that driving an older car is better for the environment, it doesn't make much sense to ignore the cars that are indeed better.
>clear history of greatly exaggerating various good news.
SpaceX and Tesla are legitimately doing great as far as I can see. Which good news were you talking about?
In the EU at least, there has been constant pressure not to do this on environmental grounds, and I believe this has been misguided (perhaps EVs will indeed change that).
> SpaceX and Tesla are legitimately doing great as far as I can see. Which good news were you talking about?
They are doing good, but they constantly promise far more - Tesla Cybertruck, Tesla Semi, "FSD" (Elon promised in ~2018 that "very soon" you would be able to let your Tesla work as a robotaxi when you don't need it and recoup its cost within a year from all the money it would make!), solar tiles.
SpaceX has similarly claimed fantastic schedules that they are far from, fantastic price cuts for orbital flight (they claim Spaceship will be at least 50x times cheaper than the cheapest option so far), that they will offer commercial earth-to-earth space flights in the near future, and that they will maintain a fleet of 40k satellites, larger than all the current number of satellites combined.
SpaceX also has a history of lying to the FAA and EPA about their launch schedules and environmental risks (not talking about CO2 here, but rocket debris in protected swamp lands).
As the saying goes: "It's not about the miles per gallon, it's about the smiles per gallon."
I see similar parallels with software. Want a sane UI with the functionality you need, or avoid being spied on? Then you have to use "insecure" older software.
Anecdotal I know, but most people I know who drive Teslas have multiple home assistants, Instagram/Facebook accounts, and little regard for personal privacy.
Most people I know in the social media/home assistant avoiding crowd talk about how they wish they could buy a car with 90s tech (but 2021 safety tech) to make it simpler and cheaper to repair, since they don't care about CarPlay or Android Auto or the million "convenience" features that don't work reliably.
If you want a high-performance engine that meets modern emissions requirements, you're probably going to be looking at EFI and turbochargers. Changing the fuel injection maps on a modern car requires something akin to rooting/flashing an Android.
It's also federally illegal, I don't know why you've veered this thread into ease of performing illegal modifications.
The closest analog with smartphones to federal emissions laws is more like the FCC prohibiting hacking the baseband ostensibly to protect the cellular network.
Contributing in small part to the electrification of transportation is more important to us than any benefit I’d enjoy from Tesla collecting less or no data from our vehicles.
In recent years there have been numerous laws and regulations passed concerning data rights and privacy. If you get consent for data usage in one way, but then use it in a way that differs with what the user consented to, that isn't in compliance with the law.
It would be fairer to say that even if the original company desperately had the desire to do so the legally required privacy compliance officer would reject their ability to do so. This is especially true at the bigger tech companies wherein they have things like external audits of their privacy compliance policies.
So why is it that so much of Hacker News as a community seems convinced that things aren't like this?
In the book Factfulness by Hans Rosling he talks about ten reasons that people get things wrong. One of them is that people assume that things stay the same rather than changing over time. One of the examples that he used for this is the widely held incorrect belief of a sharp distinction between third world and western countries. While there was once a sharp divide, the actual disparity greatly diminished. However, the legacy of that disparity continued to exist in the minds of people who weren't aware of the shifting distributions. As a result of these sort of shifting distributions one of the things he encourages in the book is regularly refreshing knowledge on a subject. I think something similar is happening here. A kind of nhilistic worldview in which things don't improve leads to people not updating their understanding of the world that changed in response to their frustration.
Even if the company itself followed the law completely, there's nothing technically stopping a bad actor (internal or external) from doing whatever they want with it if they access it.
The car they want sounds like less well made 2021 car with 1990s fuel economy and performance and a drivetrain that is easier to maintain but needs more maintenance.
Many brands have also seen a big decline in reliability and quality of craftsmanship. In fact Tesla's high-end models still don't match the quality control of Toyota's most basic cars. Sure tech has improved but construction and longevity has taken a hit.
The recurring theme is just how terrible high tech people are at estimating what’s normal in the world.
The overwhelming majority of people don’t think or care about the privacy concerns of either. The overwhelming majority of Tesla drivers are not privacy oriented tech nerds.
We are learning that humanity can learn a lot about how to treat diseases with enough understanding (data).
It’s entirely foreseeable that level 5 FSD isn’t possible without data.
How does this give anyone more data on me than is already out there based on the cellphone in my pocket and the use of my credit/debit cards? Even if Tesla didn't collect it the LTE signal of the car is enough to track me when I leave my phone at home.
And if you don't have Tesla and your car has something like OnStar it can do this too.
All I see are people making a mountain out of a mole hill. The data on my accelerator pedal is not a threat to my privacy or freedom.
The massive number of safety features in this car are a net benefit for my physical safety, though. The number of times I've avoided collisions because the car took action when people I couldn't see were about to hit me is worth every penny.
I will never buy another car without the Tesla standard set of features ever again. I used to avoid doing things because of a long drive as I hate the fatigue. Now I don't even blink at the idea of going 300 miles for an event.
Those tell Tesla nothing about my personal habits, my political leanings, who my friends are, what organizations I donate to or participate in, etc.
If the data is distinct enough, it could be used to identify the driver (with some degree of probability). It then could be shared with car insureres / law enforcement agencies / private investigators / random concerned citizens.
Tesla will just know about your driving habits and probably won't be able to sell anything based on that. It's a huge difference. Law enforcement, now that's a different story. You'll probably want to keep the privacy laws of your country strict and not allow peeking into the memory or recording of your car too easily.
That's easy. The driver sets the autopilot follow distance.
When you engage the adaptive cruise control on any car brand, you are still very much responsible to not crash in the front car.
(Obviously I'm excluding all the non-street-legal "EVs" like golf carts and the like, as well as the truly antique and rare EVs that did exist around a century ago: https://en.wikipedia.org/wiki/Category:Electric_vehicles_int... )
Fact 1: Tesla is attempting to develop a self-driving system that is based on machine learning.
Fact 2: Training machine learning systems requires training data. Having large amounts of real training data is highly desirable.
So a single step?
I'm sure that Tesla also benefits from being able to defend against false claims that would tarnish their image[1], but please don't be disingenuous about their machine learning goals.
[1] https://www.washingtonpost.com/transportation/2021/01/08/tes...
Tesla is trying to fix left turn regressions [1] in an if statement it introduced with its latest over the air update. Working on getting an iteratively better lane / cruise control assist is something many auto companies are doing, and that also doesn't make them AI companies.
1 - https://twitter.com/elonmusk/status/1451907229796417542
Edit:
By the way, here's Elon re: machine learning and Tesla.
No data connections, no logging, as hackable as possible
Maybe then it's become a game of gutting the modem (sadly).
We had this on The Record half a day before them: https://therecord.media/dutch-forensic-lab-decrypts-teslas-d...
modern-day journalism, smh
Not saying this is actually illegal, just if it is then this is probably the reasoning. One could design a drinking game around illegal things that just nobody bothers to enforce. Related: https://en.m.wikipedia.org/wiki/Possession_is_nine-tenths_of... (basically it's up to the victim to bother with legal work and proving things)