Could Volkswagen’s problems presage the ‘Internet of Cheating Things’?
The software trouble at Volkswagen has resulted in the CEO stepping down. Of course, software problems are not the domain of any one company or industry—see, for example, “EE Live keynoter: killer apps leave 30 dead,” which relates a 2014 presentation by Michael Barr, an embedded-systems expert who has served as an expert witness in product liability cases and has examined software on products ranging from automobiles to medical devices.
One would assume that most software bugs are accidental, although companies may be slow to identify and fix problems and may resort to litigation to limit liability. One would assume that engineers at reputable companies do not set out to write malware or otherwise deliberately design defective products. Of course, that’s not exactly been the case.
And now, suggests Zeynep Tufekci in the New York Times, we have entered the “era of cheating software.”
Tufekci, an assistant professor at the School of Information and Library Science at the University of North Carolina, writes, “As the Volkswagen case demonstrates, a smart object can lie and cheat. It can tell when it’s being tested, and it can beat the test.” And cheating software will have many more places from which to operate as the Internet of Things explodes—potentially becoming, as Tufekci puts it, “the Internet of Cheating Things.”
Tufekci calls for some form of regulation or oversight akin to building codes. She relates walking in Turkey in 1999 after an earthquake killed 17,000 people. Some buildings remained standing, she writes, but many others collapsed because their builders cheated on codes for concrete—using too much sand or omitting rebar. However, she writes, “Cheating software does not generate a trail of dust the way cheating concrete does.”
She offers several recommendations:
- Test smart objects “in the wild”—not just in the lab. Test cars on the highway, and parallel-test selected voting machines on Election Day.
- Don’t allow manufacturers to hide behind copyright claims. Software can be evaluated by commissions under regulatory control (or as Barr put it in 2014, it can be evaluated in clean-room environments).
- Examine what software is doing by examining its outputs—creating auditable, hard-to-tamper-with logs that regulators can inspect.
She cites class of smart devices for which these recommendations have been successfully implemented: the slot machine.
“It’s a pity that casinos have better scrutiny of their software than the code running our voting machines, cars and many other vital objects, including medical devices and even our infrastructure,” she concludes. “As computation spreads in society, our regulatory systems need to be funded appropriately and updated in their methods so that keeping our air clean and our elections honest is not a worse gamble than a slot machine.”
