August 17, 2018

Economics of Eavesdropping For Pay

Following up on Andrew’s post about eavesdropping as a profit center for telecom companies, let’s take a quick look at the economics of eavesdropping for money. We’ll assume for the sake of argument that (1) telecom (i.e. transporting bits) is a commodity so competition forces providers to sell it essentially at cost, (2) the government wants to engage in certain eavesdropping and/or data mining that requires cooperation from telecom providers, (3) cooperation is optional for each provider, and (4) the government is willing to pay providers to cooperate.

A few caveats are in order. First, we’re not talking about situations, such as traditional law enforcement eavesdropping pursuant to a warrant, where the provider is compelled to cooperate. Providers will cooperate in those situations, as they should. We’re only talking about additional eavesdropping where the providers can choose whether to cooperate. Second, we don’t care whether the government pays for cooperation or threatens retaliation for non-cooperation – either way the provider ends up with more money if it cooperates. Finally, we’re assuming that the hypothetical surveillance or data mining program, and the providers’ participation in it, is lawful; otherwise the law will (eventually) stop it. With those caveats out of the way, let the analysis begin.

Suppose a provider charges each customer an amount P for telecom service. The provider makes minimal profit at price P, because by assumption telecom is a commodity. The government offers to pay the provider an amount E per customer if the provider allows surveillance. The provider has two choices: accept the payment and offer service with surveillance at a price of P-E, or refuse the payment and offer reduced-surveillance service at price P. A rational provider will do whatever it thinks its customers prefer: Would typical customers rather save E, or would they rather avoid surveillance?

In this scenario, surveillance isn’t actually a profit center for the provider – the payment, if accepted, gets passed on to customers as a price discount. The provider is just an intermediary; the customers are actually deciding.

But of course the government won’t allow each customer to make an individual decision whether to allow surveillance – then the bad guys could pay extra to avoid being watched. If enough customers prefer for whatever reason to avoid surveillance (at a cost of E), then some provider will emerge to serve them. So the government will have to set E large enough that the number of customers who would refuse the payment is not large enough to support even one provider. This implies a decent-sized value for E.

But there’s another possibility. Suppose a provider claims to be refusing the payment, but secretly accepts the payment and allows surveillance of its customers. If customers fall for the lie, then the provider can change P while pocketing the government payment E. Now surveillance is a profit center for the provider, as long as customers don’t catch on.

If customers know that producers might be lying, savvy customers will discount a producer’s claim to be refusing the payments. So the premium customers are willing to pay for (claims of) avoiding surveillance will be smaller, and government can buy more surveillance more cheaply.

The incentives here get pretty interesting. Government benefits by undermining providers’ credibility, as that lowers the price government has to pay for surveillance. Providers who are cooperating with the government want to undermine their fellow providers’ credibility, thereby making customers less likely to buy from surveillance-resisting providers. Providers who claim, truthfully or not, to be be refusing surveillance want to pick fights with the government, making it look less likely that they’re cooperating with the government on surveillance.

If government wants to use surveillance, why doesn’t it require providers to cooperate? That’s a political question that deserves a post of its own.

Eavesdropping as a Telecom Profit Center

In 1980 AT&T was a powerful institution with a lucrative monopoly on transporting long-distance voice communications, but forbidden by law from permitting the government to eavesdrop without a warrant. Then in 1981 Judge Greene took its voice monopoly away, and in the 1980s and 90s the Internet ate the rest of its lunch. By 1996, Nicholas Negroponte wrote what many others also foresaw: “Shipping bits will be a crummy business. Transporting voice will be even worse. By 2020 … competition will render bandwidth a commodity of the worst kind, with no margins and no real basis for charging anything.

During the 1980s and 90s, AT&T cleverly got out of any business except shipping commodity bits: in 1981 it (was forced to) split off its regional phone companies; in 1996 it (voluntarily) split off its equipment-making arm as Lucent Technologies; in 2000-2001 it sold off its Wireless division to raise cash. Now AT&T long-distance bit-shipping is just a division of the former SBC, renamed AT&T.

What profit centers are left in shipping commodity bits? The United States Government spends 44 billion dollars a year on its spy agencies. It’s very plausible that the NSA is willing to pay $100 million or more for a phone/internet company to install a secret room where the NSA can spy on all the communications that pass through. A lawsuit by the EFF alleges such a room, and its existence was implicitly confirmed by the Director of National Intelligence in an interview with the El Paso Times. We know the NSA spends at least $200 million a year on information-technology outsourcing and some of this goes to phone companies such as Verizon.

Therefore, if it’s true that AT&T has such a secret room, then it may be simply that this is the only way AT&T knows how to make money off of shipping bits: it sells to the government all the information that passes through. Furthermore, economics tells us that in a commodity market, if one vendor is able to lower its price below cost, then other vendors must follow unless they also are able to make up the difference somehow. That is, there will be substantial economic pressure on all the other telecoms to accept the government’s money in exchange for access to everybody’s mail, Google searches, and phone calls.

In the end, it could be that the phone companies that cooperated with the NSA did so not for reasons of patriotism, or because their arms were twisted, but because the NSA came with a checkbook. Taking the NSA’s money may be the only remaining profit center in bit-shipping.

AT&T Explains Guilt by Association

According to government documents studied by The New York Times, the FBI asked several phone companies to analyze phone-call patterns of Americans using a technology called “communities of interest”. Verizon refused, saying that it didn’t have any such technology. AT&T, famously, did not refuse.

What is the “communities of interest” technology? It’s spelled out very clearly in a 2001 research paper from AT&T itself, entitled “Communities of Interest” (by C. Cortes, D. Pregibon, and C. Volinsky). They use high-tech data-mining algorithms to scan through the huge daily logs of every call made on the AT&T network; then they use sophisticated algorithms to analyze the connections between phone numbers: who is talking to whom? The paper literally uses the term “Guilt by Association” to describe what they’re looking for: what phone numbers are in contact with other numbers that are in contact with the bad guys?

When this research was done, back in the last century, the bad guys where people who wanted to rip off AT&T by making fraudulent credit-card calls. (Remember, back in the last century, intercontinental long-distance voice communication actually cost money!) But it’s easy to see how the FBI could use this to chase down anyone who talked to anyone who talked to a terrorist. Or even to a “terrorist.”

Here are a couple of representative diagrams from the paper:

Fig. 4. Guilt by association – what is the shortest path to a fraudulent node?

Fig. 5. A guilt by association plot. Circular nodes correspond to wireless service accounts while rectangular nodes are conventional land line accounts. Shaded nodes have been previously labeled as fraudulent by network security associates.