Features

net.wars: Shoot the cryptographer

Spurred by a horrifying case of a bank stealing its customers' lives, Wendy investigates the impossibility of censoring the Internet ...

I just finished writing some 8,000 words on censorship of the Internet, in which I said basically the things we all know: 1) it's hard; 2) any time you try you may be defeated by a bunch of rebellious nerds who feel it's their moral duty to repost the material you're trying to censor; 3) corporations are as big a threat as governments are; 4) there's technology on both sides.

And along comes a perfect example. It seems that a couple of shopkeepers in Durban named Anil and Vanitha Singh who had never been in the UK except to change planes at Heathrow woke up one day in March 2000 to find that their Diner's Club card had been used to withdraw more than £50,000 in a chain of 190 ATM transactions up and down the Piccadilly line.

Er. Right. So the bank hauled them in, took away the card, said "Lot of fraud about," the shopkeepers noted they were in South Africa at the time, and their bank agreed for a couple of weeks - and then changed its mind and sued them to recover the money. Since then, the Singhs have been investigated by the tax authorities, the Customs authorities (they import saris), and were shut down for three months while their VAT books were reviewed. Cases were drawn up. Plaintiff and defense all hire expert witnesses. Bank says, essentially, "Systems secure. Therefore, Singhs must have committed fraud."

Enter Ross Anderson (the source for this brief description of events).

Anderson has a long history of expertise on banking security and has testified in other such cases before. More than that, Michael Bond, one of Anderson's graduate students, has recently discovered a set of vulnerabilities in the hardware security modules used by most banks to protect PINs that could reduce the number of guesses a corrupt bank insider would have to make to get a particular PIN to ... 15. As the finished paper points out, such an insider could scoop up some 7,000 PINs in an average lunch break. Worth billions.

We're all supposed to trust banks. But it's commonplace in security circles that security systems are more at risk from disaffected insiders than the archetypal teenaged hacker of media glory. And unless you're going to assume that the banks are the only people in the history of the world that can create a flawless, after you eliminate the impossible, whatever remains, however improbable, must be the Dick Turpin.

What does all this have to do with censorship, I hear you cry? Well, it seems that Diner's Club South Africa (plaintiff) is administered there by Standard Bank (plaintiff), but Diner's Club internationally is owned by Citibank (plaintiff). Because Anderson, Bond, and another of Anderson's students, Richard Clayton, are all based in the UK, they were scheduled to testify in London in early March. On Wednesday, the plaintiffs asked the judge to issue an order requiring the three of them to keep all their testimony confidential.

As Anderson says in a letter to the judge, that would stop Anderson from giving expert testimony in other cases, teaching or talking about his research, and it would stop Bond from writing his PhD thesis, due this year. The hearing was yesterday, and as of this morning Anderson has not heard whether he has been enjoined from speaking or not. Besides, Anderson notes in his letter, Bond's research has already been published.

And so it has. On Wednesday, rushed out before a gag order could be imposed, in a perfect example of how hard it is to censor the Internet. Even if Anderson is now ordered to remove the paper from his Web site, there will be plenty of downloaded copies to go around. (To be fair, other parts of Bond's research had already been the subject of talks at conferences and published elsewhere; the particular piece in question is the new research discussed above.)

Anderson says that "phantom withdrawal" cases go in waves. "There was a big epidemic string of these in the early 1990s ... and there's now another epidemic of phantom withdrawals brewing. I believe it's because bank insiders have figured out how to crook your PIN. The technology for protecting them was designed in the 1970s, and it's been here for seven generations of technology. Everything else from 1974 is now in a museum or on a tip."

The problem for customers is that although they have no input into the design of the bank's security systems, they are expected to shoulder all the losses in a case of fraud. As a poster on UKCrypto said this week, the same is most certainly not true in a physical robbery, where the bank is expected to make good on any customer losses.

I think it's probably correct that customers must assume some element of the risk - such as the £50 liability on credit card losses or thefts - to ensure customers protect their cards. But it's clear from this case and this research that the balance of liabilities is wrong. Should we be allowed to know about vulnerabilities in the banking system? Or in padlocks? Or digital rights management systems?

I think we have to be. It's not the truths you know that hurt you; it's the hidden ones.