Would FPGAs have saved Volkswagen?

[ez24]

I do not really understand FPGAs.  I watch Dave's videos and read some posts related to FPGAs to try and understand them.  Right now they are black magic to me.

In one post someone says "use a FPGA because it requires less software".   Since it was software that VW used to defeat regulations (according to news reports), I was wondering if FPGAs can be more successful to defeat detection of unwanted functions. 

It seems we have a "black magic' chip (some that have 1,000s of pins and cost $39,000) in the electronics field can be used for nefarious uses because they are so complex.  Am I paranoid or not?

In other words if VW had used a FPGA would they have been caught (assume using a chip that costs less than the car)?  I assume the NSA can understand this chip (or there would be no chip) but I am thinking that even at state levels this chip cannot be understood.

Any thoughts?

[PA0PBZ]

They were not caught because someone reversed the ECU software, they were caught because real use data was so much different from the measurement data.
Nothing you can do on the hard/software side can prevent that.

[alexanderbrevig]

Nothing you can do on the hard/software side can prevent that.

Except fixing the "bug" of course

[Brutte]

I think (I suspect) that law requires proving that VW is guilty of using a "defeat device" purposely. It is not the VW responsibility to prove innocence and as the source code is a closed source, it is close to impossible to prove this was not just a software bug or an act of sabotage.

Code: [Select]

if(here()-nearest_EPA_testing_facility()>10km)
 do_not_bother_with_emission_rules();
Even removing epoxy, reading out flash content and disassembly is not a proof of their unlawful action but merely the proof of manufacturing flaw (IMHO).

[MT]

Cars today are so complicated it cost a fortune to fix anything because you need a specialcomputer to tell if a bearing is on its way
or not! In the god ol days e.g 1877 we could just listen to the noise a bearing made during turns and decide if it was time to change!   

[Rasz]

Everyone does some form of emmisions shenanigans! Someone even mentioned Honda CBR bikes on hackaday I think. Manufacturers will claim its for product longetivity - oh, we do this thing here that recirculates exhaust gases while EPA test in pro^^^^^^^^engine is cold to make it last longer!!1

[legacy]

Even removing epoxy, reading out flash content and disassembly is not a proof of their unlawful action but merely the proof of manufacturing flaw

Eh, Eh, nothing else we can say
sadly the Truth.

[vodka]

Always a Judge may sign an order for forcing to corporation to show your firmware or allow to make  reverse engineering

[tggzzz]

The problem appears to be organisational, not technical. Thus a technical "solution" would not have changed anything.

In respect of this topic, there is no conceptual difference whatsoever in using C to program bits in memory that control a processor versus using VHDL to program bits in memory that control gates.

So no, FPGAs would not have helped.

[Gyro]

I think (I suspect) that law requires proving that VW is guilty of using a "defeat device" purposely. It is not the VW responsibility to prove innocence and as the source code is a closed source, it is close to impossible to prove this was not just a software bug or an act of sabotage.

According to the BBC news this morning VW's US boss appears to have admitted the company was dishonest with the regulators, so reverse engineering is probably a moot point. 

http://www.bbc.co.uk/news/business-34322016

[Brutte]

I can understand that the closed software can or even should include some sort of „defeating device” as the trick improves the sales (their product takes advantage over competing companies). This is not a new subject and as EPA is so stupid they believe closed software has morality then that is their problem, not VW's.

The „defeating device” is unlawful only when:
 a) revealed and
 b) proved/admitted it was installed purposely.
And somehow VW managed screw up both cases 

 a) The EPA standards impose strict measuring methodology, it is only a matter of sophistication of the software that we would learn about Dieselgate close never and not in 2015. Was it that hard to write a decent „defeating device”, Mr. Volkswagen?
 b) „this kind of behavior (..) is completely inconsistent with our core values.” ??
When people are caught with pants down, they scapegoat others to avoid responsibility, Mr. Horn.. Was it that difficult to blame a buggy software, phase of the moon or Muslims? DOH..

I do not understand the VW action as it seems irrational. The risk they took was unproportional to the advantage they gained. They are not stupid, aren't they?

So what is the whole point of a confession?
The only rational idea that comes to my mind is that VW wants to avoid the EPA order of revealing the code and binary (+ compilation environment settings) at all costs. If it contained some other nasty tricks or unlicensed code, that would have grounded all their fleet, not only TDI 2.0

[Scrts]

It might be true, that some underwater agreements are going now... Maybe CEO was pushed to admit this software "feature", but save the other hidden things. On the other hand, from the law point of perspective it would be completely obvious to issue a subpoena for VW to reveal everything and do independent expert analysis on the code to check for other hidden things.

Let the heads roll down...

[legacy]

In respect of this topic, there is no conceptual difference whatsoever in using C to program bits in memory that control a processor versus using VHDL to program bits in memory that control gates.

but you can reverse the ROM content into machine code (e.g. through IDA, which is a superlative disassembler), while you can't reverse a bitstream
VHDL or Verilog is beyond the RTL, reversing a bitstream you get the fpga interconnect map, which is also chip specific (vendor::family::product), no public information available, every chip has its own map

CPU: the ISA is public, you can read opcodes from the datasheet
FPGA: just the synthetizer knows the map, the datasheet doesn't

[BradC]

Everyone does some form of emmisions shenanigans!

This is the thing. Since 1973 it has been happening and it will continue to happen. Every time someone gets caught, everyone else modifies the way they do what they do. The problem in this case is the prescriptive test environment being easy to detect and work around. It has always been this way and the only way VW got caught is someone decided to run some tests outside of the regulatory test framework. Let the arms race continue.

[GNU_Ninja]

I think (I suspect) that law requires proving that VW is guilty of using a "defeat device" purposely. It is not the VW responsibility to prove innocence and as the source code is a closed source, it is close to impossible to prove this was not just a software bug or an act of sabotage.

According to the BBC news this morning VW's US boss appears to have admitted the company was dishonest with the regulators, so reverse engineering is probably a moot point. 

http://www.bbc.co.uk/news/business-34322016

A quote from that BBC article:

It's way too early to say just yet. But the pressure is now on the car industry to prove that cheating the pollution figures isn't a widespread problem stretching across both sides of the Atlantic.

Maybe it's not just VW USA that should be worried. This could bite the ENTIRE car industry in the ass.

[tggzzz]

Maybe it's not just VW USA that should be worried. This could bite the ENTIRE global car industry in the ass. 

Oh, yes indeed. That could be very entertaining, but unfortunately we won't see that play out in public.

Another example is that to get the claimed mpg figures, some carmakers remove wing mirrors and tape up all gaps around doors (to reduce the drag).

There have been many cases in the computer industry where compiler/hardware manufacturers have been caught detecting that they are compiling a standard benchmark, and have emitted/run unrealistically well-optimised code.

[Throy]

Another example is that to get the claimed mpg figures, some carmakers remove wing mirrors and tape up all gaps around doors (to reduce the drag).

I've yet to see a car obtain the MPG that is advertised.

[GNU_Ninja]

I've yet to see a car obtain the MPG that is advertised.

Maybe the car industry should contact Batteriser (http://batteriser.com/) 800% 80% more MPG 

[Throy]

Maybe the car industry should contact Batteriser (http://batteriser.com/) 800% 80% more MPG 

[Gyro]

Maybe it's not just VW USA that should be worried. This could bite the ENTIRE car industry in the ass.

I certainly hope so, I'd be a bit upset if my car's manufacturer hadn't been trying as hard as VW to maximize my driving enjoyment while still meeting the emissions spec.   

[vodka]

Perhaps the sanction to Volkswagen Groups have been a Vendeta by General Motor ,for having broken the omerta in Spain.

i go to explain, Volkswagen revealed to Spanish Agency Competency that General Motor and others automotive corporations pacted the car prices at the  spanish trade.

And this agency imposed to all corporations (less a Volkswagen by snitch ,although they participated) a penalty of 171 Million euros .

The most affected were General Motor with 23.87 Millions € and Ford with 20.23 Millions €.

http://economia.elpais.com/economia/2015/07/28/actualidad/1438087674_837381.html

[jancumps]

Haha. Chrysler had to pay a 266€ fine in Spain.

[miguelvp]

Haha. Chrysler had to pay a 266€ fine in Spain.

Probably they forgot to turn some paperwork in time or something stupid like that.

[hamster_nz]

In respect of this topic, there is no conceptual difference whatsoever in using C to program bits in memory that control a processor versus using VHDL to program bits in memory that control gates.

but you can reverse the ROM content into machine code (e.g. through IDA, which is a superlative disassembler), while you can't reverse a bitstream
VHDL or Verilog is beyond the RTL, reversing a bitstream you get the fpga interconnect map, which is also chip specific (vendor::family::product), no public information available, every chip has its own map

CPU: the ISA is public, you can read opcodes from the datasheet
FPGA: just the synthetizer knows the map, the datasheet doesn't

Most FPGAs have bitstream encryption options too...

[linux-works]

benchmarked pc hardware also has a history of 'playing games' (lol) with the numbers.

way back in the pci video card days, I remember nvidia and matrox playing some kind of benchmark detection games with their video cards.  ethernet cards and scsi cards also could be made to detect when a benchmark is running and then change logic so that an artifical fast-path was routed thru.

if a corporation can cheat, it will.  welcome to how raw, unfettered capitalism works ;(   aka, "all's fair as long as you don't get caught".

pathetic, but then again, I stopped trusting large corporations a long LONG time ago.