Internet of criminal things
Internet of criminal things
Posted Sep 24, 2015 21:15 UTC (Thu) by dlang (guest, #313)In reply to: Internet of criminal things by pr1268
Parent article: The Internet of criminal things
The EPA has been under fire for the last month due to the massive pollution that they caused, and so have been looking for something to distract the press, the late friday release of this announcement is suspicious.
I don't think we have the full story on this.
If they made the car perform differently when it's on a dyno, it would affect the EPA smog tests, but it would also affect the performance numbers that result.
If they made the car optimize things under constant throttle, that is a valid thing to do, even on the open road.
There are a lot of accusations going around, with very little actual data.
If they programmed it to detect the specific profile of actions for the test, this would fail in a few years when the cars are being re-tested (and have to still pass emissions) because the test profile changes over time. If they were to end up with lots of their cars unable to pass the California Smog tests because the test changed slightly and no longer triggered their 'low emissions' mode, they would end up being faced with either a massive recall or class-action lawsuit from the vehicle owners.
This isn't a new engine that's just been introduced either, this is a model they've been selling and improving for several years, so they've seen these tests change over time.
So I really doubt that it's as simple a situation as it's being made out to be.
Posted Sep 24, 2015 21:36 UTC (Thu)
by jzbiciak (guest, #5246)
[Link] (7 responses)
Here's the infographic I saw that laid out which parameters the "defeat device" used to detect it was in testing mode. I don't know the source for this level of detail, so salt to taste. For those who don't care to click on the link, it looked at four factors: Position of steering, speed, duration of engine operation, and barometric pressure. That doesn't sound like how you typically optimize for constant throttle. The more common optimization is "closed loop operation," when the ECU tries to maximize fuel efficiency by straddling the narrow-band O2 sensor's threshold line.
Posted Sep 24, 2015 21:49 UTC (Thu)
by dlang (guest, #313)
[Link] (6 responses)
Which would actually make the test far more representative of the real world.
If it really is as trivial a set of tests as this shows (again, unknown source, unknown reliability), just wiggling the steering wheel during a test would result in drastic power and emissions changes, with no throttle changes at all.
I know that the initial findings of this were based on doing emissions testing on the road, not on a dyno, resulting in drastically different emissions. But from that it's not necessarily even anything wrong with the car, but rather something wrong with the test (the purpose of the test is to try and simulate real-world driving)
It also wouldn't be the first time that the EPA test was found to be so horrifically different from real driving that it needed to be changed. As I've said before (Elsewhere at least), the early Hybrid cars resulted in insanely good test results because the profile let them run on battery most of the time.
It may just be time for the EPA to make a drastic (not just incrimental) change to their test process.
Very little real-world driving involves constant throttle settings over any significant distance, with modern computer controlled Dynos, they should be able to do tests where they vary the throttle, vary the load, vary the speed, etc.
Even the systems in Smog Test stations around the country allow for many different test profiles.
Posted Sep 24, 2015 22:25 UTC (Thu)
by dlang (guest, #313)
[Link] (5 responses)
if you have one set of wheels stationary and the other set moving, that's going to trigger your traction problem detection and attempt to slow the spinning wheels.
Apparently this is even more interesting on the very high-end cars, and so to test those cars you have to do some specific things to disable the traction control. For VW to detect a dyno mode to disable things like this is a user-friendly and mechanic-friendly thing to do.
Since you don't have the same airflow and cooling on the dyno (for things like your tranmission, differentials, etc) there's also legitimate reasons to be more cautions about thinks that can generate heat in those areas.
Again, we really need to see the results of the investigation.
Posted Sep 24, 2015 22:54 UTC (Thu)
by jzbiciak (guest, #5246)
[Link] (4 responses)
I don't think it's specifically dyno vs. not-dyno that's at play here. Given the number of people who like to tune their cars, soup them up, etc., you'd think someone would notice messed up power curves in a dyno test. My guess (and yes, it's just pure speculation) is that the code actually looked very narrowly for the test profile, and that the EPA test profile used for certification is fairly fixed. You're right, though: We really do need to see the results of the investigation. I'm quite curious what exactly they did. Knowing how these things go, we may not know the specifics for years, though.
Posted Sep 24, 2015 23:23 UTC (Thu)
by dlang (guest, #313)
[Link] (3 responses)
The EPA test profile is very fixed, but the profile used at Smog Check stations in California (where the strictest emissions requirements are), both varies over time and is subject to a lot more variation than the "official EPA" test, both from the impossibility of keeping so many thousands of stations _exactly_ in tune, and the fact that they all run their tests on ambient air, rather than 'Standard Temp and Pressure' the way the official EPA test does. The fact that the cars going through the periodic testing are going to have wildly different internal drag (tire pressure, how fresh the lubricants throughout the vehicle are), is going to make it so that the throttle setting needed to run the car at a specific speed for the test is going to vary a fair bit.
This is why I'm a bit sceptical that this is deliberate cheating. The fact that California can and does change it's test profile FAR more frequently than the EPA does makes detecting specific driving profiles much harder.
If (as reported) not triggering the 'EPA test mode" made the NOX levels exceed the testing limits by 40x, just can't see how these cars have been triggering 'test mode' reliably enough to not be an epidemic of failures in the California Smog Test stations.
I'm not saying it's impossible, but some things are far more likely than others, and I know enough of the field (having been Smog Certified in California as well as a car performance guy, and then Information Security as my day job) that some things are far more likely than others, and these claims are odd enough for me to question them. Or at least question that we have valid info yet.
Posted Sep 24, 2015 23:34 UTC (Thu)
by jzbiciak (guest, #5246)
[Link]
Gotcha. I guess I'm just going to have to stay tuned.
Posted Sep 25, 2015 13:40 UTC (Fri)
by mstone_ (subscriber, #66309)
[Link] (1 responses)
The allegation is that VW cheated on the initial EPA test, so that all of the subsequent reporting (including everything done by CA) is invalid.
Posted Sep 25, 2015 17:05 UTC (Fri)
by dlang (guest, #313)
[Link]
When I was first certified, the Dyno was not used for any vehicles, it got added much more recently.
If there has not been an update to the check requirements for Diesels to have them use the same Dyno profiles and emissions sniffing process as Gas vehicles do, it's only a matter of time until there is (and after this mess, I would expect that it will be a fairly short time)
Posted Sep 25, 2015 7:43 UTC (Fri)
by pr1268 (guest, #24648)
[Link] (10 responses)
Sure we do. VW intentionally cheated the emissions tests with a so-called "defeat device". This investigation has been going on for almost two years (according to our editor's PDF link). The $**t hit the fan only a few days ago when VW came clean (no pun intended) and admitted their scheme. In fact, this article from last Friday (Sept. 18) said that 500,000 vehicles were affected. But, VW then later admitted 11 million cars worldwide were rigged. Sounds pretty cut-and-dry, IMO. I'm sure the EPA is relieved at how the VW scandal has deflected attention away from the polluted Animas River, but from what I can tell, the VW scandal had been brewing for quite some time prior to the mine leak. No one was accusing anyone of anything, other than the EPA threatening to withhold certification of VW's 2016 model year diesel cars until the emissions could be fixed. Only then did VW make a public admission of guilt to using software to cheat the emissions tests.
Posted Sep 27, 2015 11:36 UTC (Sun)
by ballombe (subscriber, #9523)
[Link] (9 responses)
> Only then did VW make a public admission of guilt to using software to cheat the emissions test
This is true, but this is still very awkward: US companies never make public admission of guilt under any circumstance, they settle the case with the government without admitting wrongdoing.
Posted Sep 27, 2015 12:33 UTC (Sun)
by pr1268 (guest, #24648)
[Link] (8 responses)
I honestly don't know. I was sort of begging the same question with my previous post(s) here on LWN. I suppose it might be a cultural difference between Germany and the USA. Or a legal one. A slightly whacky analogy I draw VW's actions to is that of a murder trial: Just as the prosecution is about to prove the defendant guilty, the defendant then proclaims that not only did he commit the murder, but he murdered ten others, and here's where the bodies are buried! :-\
Posted Sep 27, 2015 22:06 UTC (Sun)
by raven667 (subscriber, #5198)
[Link] (7 responses)
The cynical person would say that maybe it's like Watergate where the actions taken were to hide a much more serious crime than what the public knew about at the time.
Posted Sep 28, 2015 19:21 UTC (Mon)
by bronson (subscriber, #4806)
[Link] (6 responses)
It took a lot of work and fact checking by CARB/EPA/etc to wear them down:
"We discovered some very strange anomalies," Young said. "For instance, the car was running more cleanly when it was cold than when it was warm, which is the opposite of what every other car does — because once you warm a car up that's when it begins to deliver its best pollution controls. This was not the case. So clearly something else was going on. Over time we assembled enough proof and questions that they could no longer provide any reasonable explanation for what was going on."
I think VW just ran out of things they could plausibly lie about.
Also, pure speculation: engine computers are pretty standardized and not too hard to analyze (much easier than a locked-down smartphone anyway!) It wouldn't surprise me if CARB found evidence of a defeat device on their own and quietly confronted VW about it.
Posted Sep 28, 2015 19:45 UTC (Mon)
by raven667 (subscriber, #5198)
[Link]
True, but that has not stopped many other companies before from continuing to lie, even to try propaganda and lobbyists to change the law and public opinion to support their untruth, so VW is actually different in this case.
Posted Sep 28, 2015 19:46 UTC (Mon)
by dlang (guest, #313)
[Link] (4 responses)
Posted Sep 28, 2015 19:50 UTC (Mon)
by bronson (subscriber, #4806)
[Link] (3 responses)
But, yes, we're probably talking about a few tens or hundreds of lines of software. It's the EPA that named it a "defeat device", not me. :)
Why is it worth repeating?
Posted Sep 28, 2015 20:15 UTC (Mon)
by dlang (guest, #313)
[Link] (2 responses)
Not just cars, but the FCC is currently accepting comments on proposing requiring that access point manufacturers be required to demonstrate how they will prevent people who buy the devices from installing DD-WRT or OpenWRT on the devices based on the fact that such open software is a "defeat device"
we lost the definition of "hacker", don't let if statements (and similar code) start being talked about as if it was a hardware device installed in something that can just be locked out.
Posted Sep 28, 2015 20:46 UTC (Mon)
by bronson (subscriber, #4806)
[Link] (1 responses)
However, "defeat device" is a term coined in the 70s (if not before) when the government was trying to outlaw physical devices designed to defeat emissions testing. Law (as it does) carried that terminology forward into the era of the engine computer: https://www.law.cornell.edu/cfr/text/40/86.1809-10
So, even if you find it to be scare words, that's not the way it was originally intended. I think it can be forgiven in this case.
Just curious, can you suggest a similarly unambiguous term that can be codified into law? Remember, the term needs to cover both software and hardware.
Posted Sep 28, 2015 21:41 UTC (Mon)
by dlang (guest, #313)
[Link]
But I've seen people elsewhere start getting up in arms about the conspiracy because this device has been installed in all these cars for years and nobody has spotted it.
When you point out that it's not a physical device, it's just software in the system, the expectations change. There's still plenty of silly, over-the-top reactions even then.
I don't know a good phrase to try and replace it as a "term of art" in legal matters though :-(
Internet of criminal things
Internet of criminal things
Internet of criminal things
Internet of criminal things
Internet of criminal things
Internet of criminal things
Internet of criminal things
Internet of criminal things
Internet of criminal things
personally I don't think we know for sure what VW did. The execs agreed to cooperate with the investigation this last weekend, which means that there still needs to be an investigation.
The EPA has been under fire for the last month due to the massive pollution that they caused, and so have been looking for something to distract the press, the late friday release of this announcement is suspicious.
There are a lot of accusations going around, with very little actual data.
Internet of criminal things
Why is VW doing otherwise, when it is so easy to say it is just a software bug?
Internet of criminal things
Why is VW doing otherwise, when it is so easy to say it is just a software bug?
Internet of criminal things
Internet of criminal things
Internet of criminal things
Internet of criminal things
Internet of criminal things
Internet of criminal things
Internet of criminal things
Internet of criminal things
