Authorization

A secret tool to catch the next VW-style emissions cheat

A secret tool to catch the next VW-style emissions cheatVolkswagen’s vast diesel emissions fraud went undetected for years. To finally expose it, an academic group attached portable emissions measurement systems to a diesel Passat and Jetta, then drove the cars round-trip from Seattle to Los Angeles to directly measure exhaust in real-world conditions and prove that VW had programmed the cars to disable the controls.

New software under development could do the same thing in just hours by monitoring vehicle data to infer problems with emissions controls without any need to sample the exhaust or directly detect that emissions controls weren’t engaged.

“It definitely could have caught the VW cheating earlier. If there was a way for emissions enforcement authorities to install this software in cars, we could detect it automatically,” says Armin Wasicek, a postdoc at the University of California, Berkeley, who led the development of the software analysis tool.

His technical paper on the topic, co-produced with a group of researchers at the University of Michigan’s Transportation Research Institute and pending publication, builds on earlier work on the concept. The new research was shown to detect “chip-tuning”—wherein engine control chips are altered to boost horsepower (a trick that tends to worsen emissions). The same detection method would also pick up disabling of emissions controls.

In the VW scandal, diesel emissions controls were designed to work only when the car was stationary and the testing equipment was plugged in. When the car left the testbed and hit the road, proprietary VW software—out of the view or control of regulators—would disable the system. The result: peppier performance and far worse emissions for 500,000 cars on U.S. roads and perhaps 11 million worldwide. Similar scandals are now emerging at Mitsubishi and perhaps other carmakers.

Detecting hacks of engine chips or disabled emissions controls doesn’t necessarily require sniffing the exhaust. Reams of data from the car—on the torque produced by the engine, the engine’s revolutions per minute, the rate of fuel consumption, the temperature of the catalytic converter, and even the position of the accelerator pedal—can, when taken together, provide a telltale fingerprint.

Read also: China’s CO2 emissions might not be falling – Study

And that’s what the new software does. First it captures data on what a given car’s operating parameters look like during normal operations, then compares it to what’s actually happening. The new technology was able to detect a hacked engine chip when tested in a real car, but the concept could be broadly applicable.

“We could put this inside the car’s computer system and automate the whole thing,” Wasicek says. Then reports on anomalies and their potential causes could be streamed from the cars or stored for later downloading.

The long-term vision is that manufacturers, insurers, registration agencies, and emissions authorities could keep an eye on compliance, safety, and performance—subject to resolving privacy concerns, he says.

The work adds to a large body of security research in the automotive industry in the wake of high-profile hacking demonstrations in the past two years. Car companies are stepping up their security efforts, and big security companies like Symantec are also working on concepts for detecting problems with vehicle IT systems—whether malicious hacks or disabled emissions controls.

Shankar Somasundaram, senior director of Internet of things at Symantec, says that Wasicek’s software is making a key contribution to the field. “Future-looking work like [his] on automotive intrusion detection is so important,” says Somasundaram. “It provides us alternative ways to look at a vital need as the industry continues to evolve.”

Source: MIT Technology Review
.
See also:
Leave a comment
TOP Video
News
  • Latest
  • Read
  • Commented
Calendar Content
«    Декабрь 2016    »
ПнВтСрЧтПтСбВс
 1234
567891011
12131415161718
19202122232425
262728293031