Freedom to Tinker

Advertisement

Close

This week the European Commission (EC) announced that it is fining Google $2.7 billion for anti-competitive tactics in the company’s iconic search product. In this post I’ll unpack what’s going on here.

I have some background on this topic. In 2011-12, when I was Chief Technologist at the FTC, the agency did a big investigation on this same topic. The FTC eventually decided not to bring a case against Google for this behavior. The EC has now reached a different conclusion.

The EC makes two main claims. First, they claim that Google dominates the search engine market in Europe–it’s pretty hard to argue with that.  Second, they claim Google designed its dominant search product in ways that unfairly advantage the company’s own Google Shopping product and unfairly disadvantage competing comparison shopping products.

Competition law is complicated, and I won’t presume to offer any legal analysis. But the basic principles motivating competition policy are not too complicated. Fair competition is encouraged. If your business grows because you improve your product, or manage your operations well, or negotiate shrewdly, or simply happen to be in the right place at the right time, that’s all good. If you amass dump trucks full of money doing this, then good for you, and thank you for your tax dollars. That’s how capitalism is supposed to work.

But if your effort is devoted to preventing fair competition, then you are probably harming consumers, and that’s a competition policy problem.  To see the difference, suppose you’re in the business of delivering packages to people’s homes.  Fair competition means buying better trucks, optimizing routes and schedules, hiring better employees, and so on.  But if you send out employees to block your competitors’ trucks, that is an anticompetitive tactic.

Now back to Google. The EC says that when users do searches relevant to shopping, Google gives its own Google Shopping product preferred placement in the search results–and higher placement leads to more clicks and more sales–while demoting competing shopping products in the search results. These two claims, self-promotion and competitor-demotion, may sound similar at first, but they raise different issues for us in understanding the case, so let’s look at them separately.

On the self-promotion claim, we know the relevant facts. On shopping-relevant searches, Google puts a box at or near the top of the search results, showing Google Shopping results with images of items for sale. That is a valuable benefit that Google Search is giving to the Google Shopping product. Is this anticompetitive? Google’s strongest argument to the contrary is that the Shopping box is essentially an ad, and Google already places ads at the top of the page. If Google auctioned that space off to the highest bidder for advertising, nobody would object. So why is it a problem if Google gives that advertising space to Google Shopping? The company could make a symbolic payment to itself to buy the space, if that made a difference to anybody.

The competitor-demotion claim is very different–the theory is less complicated, but the analysis depends more on facts not available to the public. If Google is gratuitously demoting its shopping competitors in search results, that is problematic. But Google says it is not doing that–it says that those competitors’ placements arise naturally from a search ranking algorithm based on design decisions that the company made for legitimate, pro-consumer reasons.

It’s hard for the public to tell who is right. Google’s ranking algorithm is complicated, and it changes constantly, as the Web changes and as Google works to counter sites’ attempts to game the algorithm. Is there evidence that Google tweaked the algorithm with the goal of demoting shopping competitors? Did the company make algorithm changes for the wrong reasons, or did suspicious changes happen outside the normal process? These questions are answerable in principle, but only by looking at the company’s internal information, which the EC might have but we, the public, do not.

At this point, I need to put some of my cards on the table and admit that I know more about this topic, having worked on the FTC’s investigation which asked some of the same questions. But that investigation was confidential, for good reasons, and I will not violate that confidentiality. All I’ll say is that the FTC had the legal power to compel answers to factual questions about Google’s practices (and an obligation to keep the answers confidential) and, having conducted a thorough investigation, the FTC decided not to bring a case against Google.

So why did the European authorities get a different result than the U.S. authorities? The answer might lie in differences between European and American competition law. Or it might lie in the fact that European authorities find it easier to enforce against a foreign company. Regardless of the reason, Google is presumably looking for ways to resolve the complaints that led to this investigation being started.

The US National Highway Traffic Safety Administration (NHTSA) is proposing a requirement that every car should broadcast a cleartext message specifying its exact position, speed, and heading ten times per second. In comments filed in April, during the 90-day comment period, we (specifically, Leo Reyzin, Anna Lysyanskaya, Vitaly Shmatikov, Adam Smith, together with the CDT via Joseph Lorenzo Hall and Joseph Jerome) argued that this requirement will result in a significant loss to privacy. Others have aptly argued that the proposed system also has serious security challenges and cannot prevent potentially deadly malicious broadcasts, and that it will be outdated before it is deployed. In this post I focus on privacy, though I think security problems and resulting safety risks are also important to consider.

The basic summary of the proposal, known as Dedicated Short Range Communication (DSRC), is as follows. From the moment a car turns on and every tenth of a second until it shuts off, it will broadcast a so-called “basic safety message” (BSM) to within a minimum distance of 300m. The message will include position (with accuracy of 1.5m), speed, heading, acceleration, yaw rate, path history for the past 300m, predicted path curvature, steering wheel angle, car length and width rounded to 20cm precision, and a few other indicators. Each message will also include a temporary vehicle id (randomly generated and changed every five minutes), to enable receivers to tell whether they are hearing from the same car or from different cars.

Under the proposal, each message will be digitally signed. Each car will be provisioned with 20 certificates (and corresponding secret keys) per week, and will cycle through these certificates during the week, using each one for five minutes at a time. Certificates will be revocable; revocation is meant to guard against incorrect (malicious or erroneous) information in the broadcast messages, though there is no concrete proposal for how to detect such incorrect information.

It is not hard to see that if such a system were to be deployed, a powerful antenna could easily listen to messages from well over the 300m design radius (we’ve seen examples of design range being extended by two or three orders of magnitude through the use of good antennas with bluetooth and wifi). Combining data from several antennas, one could easily link messages together, figuring out where each car was parked, what path it took, and where it ended up. This information will often enable one to link the car to an individual–for example, by looking at the address where the car is parked at night.

The fundamental privacy problem with the proposal is that messages can be linked together even though they have no long-term ids. The linking is simplest, of course, when the temporary id does not change, which makes it easy to track a car for five minutes. When the temporary id changes, two consecutive messages can be easily linked using the high-precision position information they contain. One also doesn’t have to observe the exact moment that the temporary id changes: it is possible to link messages by a variety of so-called “quasi-identifiers,” such as car dimensions; position in relation to other cars; the relationship between acceleration, steering wheel angle, and yaw, which will differ for different models; variability in how different models calculate path history; repeated certificates; etc. You can read more about various linking methods in our comments; and in comments by the EFF.

Thus, by using an antenna and a laptop, one could put a neighborhood under ubiquitous real-time surveillance — a boon to stalkers and burglars. Well-resourced companies, crime bosses, and government agencies could easily surveill movements of a large population in real time for pennies per car per year.

To our surprise, the NHTSA proposal did not consider the cost of lost privacy in its cost-benefit analysis; instead, it considered only “perceived” privacy loss as a cost. The adjective “perceived” in this context is a convenient way to dismiss privacy concerns as figments of imagination, despite the fact that NHTSA-commissioned analysis found that BSM-based tracking would be quite easy.

What about the safety benefits of proposed technology? Are they worth the privacy loss? As the EFF and Brad Templeton (among others) have argued, the proposed mandate will take away money from other safety technologies that are likely to have broader applications and raise fewer privacy concerns. The proposed technology is already becoming outdated, and will be even more out of date by the time it is deployed widely enough to make any difference.

But, you may object, isn’t vehicle privacy already dead? What about license plate scanners, cell-phone-based tracking, or aerial tracking from drones? Indeed, all of these technologies are a threat to vehicle privacy. None of them, however, permits tracking quite as cheaply, undetectably, and pervasively. For example, license-plate scanners require visual contact and are more conspicuous that a hidden radio antenna would be. A report commissioned by NHTSA concluded that other approaches did not seem practical for aggregate tracking.

Moreover, it is important to avoid the fallacy of relative privation: even if there are other ways of tracking cars today, we should not add one more, which will be mandated by the government for decades to come. To fix existing privacy problems, we can work on technical approaches for making cell phones harder to track or on regulatory restrictions on the use of license plate scanners. Instead of creating new privacy problems that will persist for decades, we should be working on reducing the ones that exist.

The 2016 election was one of the most eventful in U.S. history. We will be debating its consequences for a long time. For those of us who pay attention to the security and reliability of elections, the 2016 election teaches some important lessons. I’ll review some of them in this post.

First, though, let’s review what has not changed. The level of election security varies considerably from place to place in the United States, depending on management, procedures, and of course technology choices. Places that rely on paperless voting systems, such as touchscreen voting machines that record votes directly in computer memories (so-called DREs), are at higher risk, because of the malleability of computer memory and the lack of an auditable record of the vote that was seen directly by the voter. Much better are systems such as precinct-count optical scan, in which the voter marks a paper ballot and feeds the ballot through an electronic scanner, and the ballot is collected in a ballot box as a record of the vote. The advantage of such a system is that a post-election audit that compares a random sample of paper ballots to the corresponding electronic records can verify with high confidence that the election results are consistent with what voters saw. Of course, you have to make the audit a routine post-election procedure.

Now, on to the lessons of 2016.

The first lesson is that nation-state adversaries may be more aggressive than we had thought. Russia took aggressive action in advance of the 2016 U.S. election, and showed signs of preparing for an attack that would disrupt or steal the election. Fortunately they did not carry out such an attack–although they did take other actions to influence the election. In the future, we will have to assume the presence of aggressive, highly capable nation-state adversaries, which we knew to be possible in principle before, but now seem more likely.

The second lesson is that we should be paying more attention to attacks that aim to undermine the legitimacy of an election rather than changing the election’s result. Election-stealing attacks have gotten most of the attention up to now–and we are still vulnerable to them in some places–but it appears that external threat actors may be more interested in attacking legitimacy.

Attacks on legitimacy could take several forms. An attacker could disrupt the operation of the election, for example, by corrupting voter registration databases so there is uncertainty about whether the correct people were allowed to vote. They could interfere with post-election tallying processes, so that incorrect results were reported–an attack that might have the intended effect even if the results were eventually corrected. Or the attacker might fabricate evidence of an attack, and release the false evidence after the election.

Legitimacy attacks could be easier to carry out than election-stealing attacks, as well. For one thing, a legitimacy attacker will typically want the attack to be discovered, although they might want to avoid having the culprit identified. By contrast, an election-stealing attack must avoid detection in order to succeed. (If detected, it might function as a legitimacy attack.)

The good news is that steps like adopting auditable paper ballots and conducting routine post-election audits are useful against both election-stealing and legitimacy attacks. If we have strong evidence of voter intent, this will make election-stealing harder, and it will make falsified evidence of election-stealing less plausible. But attacks that aim to disrupt the election process may require different types of defenses.

One thing is certain: election workers have a very difficult job, and they need all of the help they can get, from the best technology to the best procedures, if we are going to reach the level of security we need.