Privacy Subtleties of GMail

Much has been written about the new Google GMail trial, which is an e-mail service that offers a gigabyte of archiving, Google search of your mail archives and a nice interface. It's free, because while you read your mail, Google ads will be displayed based on keywords found in the mail you are reading.

GMail created a surprising storm for a product that hasn't yet been released. A coalition of privacy groups asked Google to hold back on releasing it. A California state senator proposed a law to ban the advertising function. Editorials and blog entries left and right have condemned it and praised it.

I come to this problem from two sides. One, I'm a fan of Google, and have been friends with Google's management since they started the company. I've also done work for Google advising on software design.

I'm also a privacy advocate and Chairman of the Electronic Frontier Foundation, well regarded as one of the top civil rights advocates in cyberspace. The EFF has issued some statements of privacy concern over GMail, though we declined joining the coalition against it. (I'm writing this as my own essay, though with some advice from the EFF team.) I've also had a chance to talk at length with Google President Larry Page about some of the issues. (Update: My term as EFF chairman expired in 2010.)

Here's a summary of some of my conclusions

While there has been over-reaction to GMail, there are some real issues here to be worried about.
Other webmail providers are doing, or will be doing the same things, meaning these issues apply to all of them, including MSN, Yahoo and others.
One key risk is that because GMail gets your consent to be more than an e-mail delivery service -- offering searching, storage and shopping -- your mail there may not get the legal protection the ECPA gives you on E-mail.
The storage of e-mail on 3rd party servers for more than 180 days almost certainly causes the loss of those privileges.
This in turn creates a danger that we may redefine whether e-mail has the "reasonable expectation of privacy" needed for 4th amendment protection.
Correlation of search and mail has real risks.
Google and others should architect to encrypt your mail.
Even the irrational fears over the spooky aspect of advertising being associated with e-mail creates problems that must be addressed.

Page and the others in Google were taken aback by the negative reaction to their pre-launch. GMail is a very nice product, with great promise. They've done a lot of work on building a superior web e-mail interface. Their surprise at the tempest is not inappropriate. Many of the issues raised here are subtle and known only to full-time privacy lawyers. In addition, many of the debated features in GMail are already present to some degree in competing products like MSN Hotmail, Yahoo Mail and AOL. And indeed, some of the reaction has been silly and a bit paranoid, and some of that has gotten a lot of the attention.

But there are also some deep issues here, worth discussing with not just Google but all the other webmail providers.

The thing many people reacted to the most was the idea of displaying ads keyed to words in your mail. We think of our mail as fairly private, and we do a lot of private stuff in e-mail. Google's ad-linking is all done by computer, they promise not to have human beings look at the mail (almost all the time). If only a computer "knows" your deepest secrets, is there a concern? I certainly let my own personal computer contain all sorts of private information, ranging from my finances to private e-mails with my family.

Even so, people have a reaction to a 3rd party computer doing scans like this. If you were offered a service that saved you money by having your paper mail opened by robots for scanning, which then inserted new junk mail in your box based on what it found, you might get a bit creeped out. Go further and consider a service that gave you free phone calls if it could have speech-recognizing computers listen in and barge in with product offers related to your conversation? It's easy to imagine an unpleasant situation where you get invited to a gay wedding in Vancouver, and find with it in your mailbox brochures for gifts, Vancouver hotels and a free copy of Out magazine. People have extended that fear into the e-mail realm.

Of course, webmail is an optional service. Those afraid of these scenerios can simply not use it, and that's an entirely reasonable answer. But it ignores an aspect of the privacy fears we have. In the modern era where computers threaten privacy, we are as afraid of outside computers knowing all about our lives as much as we are outside people. Even though we might trust the people running the computers today, or the human health insurance clerk who learns we have cancer, we are uncomfortable with both of those things. We are not quite as uncomfortable about the computer knowing, but it's hard to ignore the fear that, in spite of the best of intentions, information on external computers (and even our own computers) sometimes makes it out into the world.

People's fears here are sometimes irrational, but still real. And irrational fears often affect privacy and freedom, even when they are irrational. Consider the amusing Wall Street Journal story some time ago headlined, "Help, my Tivo thinks I'm gay!" (The Tivo suggests shows to you based on what you have watched before. Tivo says that this does not leave the box associated with your identity, though they do collect data on all Tivo use, stripping off the identity along the way.) People became annoyed at or scared by their video recorders.

Google is not ignorant of this issue, of course. They plan to work hard to not do something so bold as sexual ads on your salacious e-mails with your husband. They know that even irrational freaking-out generates a real consumer issue.

And it's not to be ignored that well-targetted advertising is itself a useful thing. If you take as a given that you're going to accept ads to subsidize an activity, few wish to have their time wasted by ads that are irrelevant. In addition, well-aimed advertising costs advertisers more because it is more effective -- which means you can get the ad subsidy you are seeking with fewer ads. Clearly many users find themselves buying products from these ads, products they might not have known about without them, so good ads can be a service unto themselves.

I'll talk about the result of some of the irrational fears in a moment, but first let's look at some of the real fears.

Our lives move online and outside

GMail is a big step in a concerning direction. If it's a success, millions of people will move a great deal of the record of their lives not just online, but online and stored with a 3rd party. This isn't the first time this will happen, but for millions it will be the biggest such jump. Particularly if one's searches and social network can be correlated with this information. My e-mail contains the story of my life, and what's not in there is often recorded in my searches.

As we move these things online and outside, we build some of the apparatus for a surveillance society. As usual, we don't plan to do so, and the people building it would oppose it being used that way. But they build it nonetheless. We make it so that having that surveillance becomes a "logical" step -- changing a law or a policy, or in some cases just pushing a button, rather than a physical step -- going into somebody's house to grab their papers.

When our papers are at home, mass surveillance of them simply doesn't scale. It's too expensive. Online, it scales well. Our networked computers at home are not very secure, of course -- probably less secure than Google's against unauthorized intruders. But they must be broken into a million times -- though possibly by one giant virus -- and those outside machines must be compromised (by the law or by system crackers) only once.

Webmail is, as noted, only a part of this trend. No surprise, since there are many technical advantages to centralization, especially when it comes to software maintenance. Installing, running and upgrading software on your home computer is hard. A remote computer is professional maintained, and almost surely a lot faster too. It can be reached from anywhere. In many cases, where you want to roam all around and access your data, or have servers receive data for you while your home computer is off, it's the only way to do things well. I won't pretend that such centralization is not quite useful, or imagine it is something we can stop from happening.

Many have written, correctly, that we have already seen this trend in Hotmail, AOL, Yahoo, Turbotax Online, Instant Messengers, Salesforce.com and all other "Application Service Providers" (ASPs.) That the trend has already started doesn't make it less disturbing.

GMail however, with the offer of a free gigabyte, makes a big jump in this trend. Free Hotmail users tended to keep only a bit of their mail online, deleting the rest. Hotmail forced them to. GMail (and soon other webmail offerings) will openly encourage the opposite.

Google is a good company with honest people, headquartered in a fairly free country with protection of rights that is among the best in the world. But that's not always so. I wonder, for example, about what it means to the people of China that we have built our instant messaging infrastructure and some of our E-mail infrastructure entirely with centralized servers and no encryption? Their government can and does routinely listen to internet traffic. When we build our systems, how often do we think of what it will mean when they become popular in China, or Saudi Arabia? Or what it means when they are sold to other companies whose policies are not so benevolent.

There are ways to buck this trend. Below, I'll be recommending that Google and its competitors work to encrypt your data when they store it for you.

ECPA

The Electronic Communications Privacy Act started out in the 80s as a blessing. It declared that e-mail was a private means of communication, and that we might hope for the same level of privacy in it as we have in phone calls and letters. Among other things, it means that police need a wiretap warrant to read your e-mails, and that your e-mail company's employees can't disclose your e-mails to others.

But the world has changed and the ECPA has not changed with it. E-mail in transit is protected, but those in law enforcement advocate that once mail is processed and stored, it is no longer the same private letter, but simply a database service.

GMail's big selling point is that they don't simply deliver your mail. They store it for you, and they index it so you can search it. (Of particular importance, you authorize them to scan your mail for these purposes, and that authorization is the act that risks stripping you of your rights.) All the other webmail companies who don't already do this will, I'm sure, quickly have offerings to compete with GMail. As noted, not only do they search it, but they scan on viewing to provide ads that match the content, making it not just a database but a shopping service. Should you click on those ads, the merchants will see your IP address, and know somebody from that address (or with their cookie) was reading a page, search result or e-mail related to their ads.

Unfortunately, a database and shopping service doesn't look as much like an e-mail delivery service as it should according to the legal definitions in the ECPA. Thus, while Google promises not to peek into your e-mails or hand them to others, the danger is that this is now solely their contractual promise. With an e-mail delivery service covered by the ECPA, the law, not just Google's terms of service governs when your mail can be handed over, particularly to law enforcement.

Google also has some work to do on their privacy policy, because it does allow them to look at or release your e-mail in circumstances where they would be forbidden to do so if it were ECPA protected, such as a law enforcement "request." But even if it bound them as tightly as the law does, it still contains the ubiquitous clause letting them change the policy at will, without your consent -- so it's really not a strong restriction at all. The policy should state at a minimum that changes to it which would reduce the privacy of your mail can only take place with your explicit and informed consent at the time of the change.

A lot of privacy policies should say that, of course. They don't, because doing this in general is a logistic nightmare. However, with work, one can fine-tune such policies to dictate what can be changed and what can't, to do at least as well as what the law requires of e-mail delivery companies, if not better. This is not easy; it's hard. But it's worth doing.

When you use network services, you need a contract, and that contract will be written primarily to protect the interests of the service provider. When you run software on your PC, you may agree to terms when you install software, but as far as your own data is concerned there is usually no question of contract at all. The data is on your PC, and there don't have to be rules governing what others can do with it.

Without the ECPA protection, your e-mail (now just a database) can be seized against Google's will by the government with an ordinary subpoena (vastly less involved than a warrant or wiretap) or in the discovery phase of a lawsuit. With warrants, and in some rare cases even without them, your mail can be grabbed without you being informed that this has been done. Worse, Google has retained the right to hand it over in the case of a "request" from law enforcement, rather than a court order.

Of course, these legal techniques can be applied to the data you store yourself. However, short of a secret warrant to break into your house and seize your computer, it can't be done without your knowledge and involvement. You have the chance to fight any attempt to grab your mail. You have the power to get a lawyer and appeal any order in front of a judge. You give up some of that power when you put your e-mail database in somebody else's hands. Google is a good company that wants to please its users, they don't want to be subject to all these subpoenae. But they won't fight as hard as you would.

It's also important to note that most of these webmail providers are global companies, with servers around the world. The ECPA is a U.S. law, only protecting mail in the USA. Some other nations have protections but they certainly all don't. Unless care is taken, your mail could end up stored in another country without legal protections you were hoping for.

Now, after scaring you like this, let me add that this is an entirely new situation, and the courts have not yet ruled whether the ECPA covers your mail in a searching/advertising database service. It's possible they might, but from the text of the law, quite possible they would not. Past history suggests it is highly likely the Department of Justice will take the position that the protection is lost.

The DoJ in fact believes that the moment after you open your mail, it is just an archive and loses some protections, though at least the 9th circuit court (where Google lives) has disagreed with that -- up to the 180 day rule at least.

To learn more about the ECPA and e-mail you will find a good analysis in this paper. It's way more complex than I describe here.

The 180 day rule

In the hoped-for event that your webmail archives are fully protected by the ECPA as what it calls an ECS, they lose some of that protection after 180 days. This is not news, but a product like GMail, which encourages long-term archving of e-mail with the web mail provider brings the question to the forefront. After 180 days your e-mail archives can now be fetched by the government without a warrant, through a special ECPA court order or a subpoena. (In most cases, but not all, you will get notice of such seizures.)

Expectation of Privacy

In the USA, privacy from government intrusion is defended by the Fourth amendment, which requires a warrant be issued by a judge for many kinds of searches and seizures. Over the years, court precedents have required that warrant when you have a "reasonable expectation of privacy" over what's being searched. The definition of when and where you have this expectation of privacy has been in flux. Court decisions have sometimes sustained the expectation of privacy, but often eroded it.

The higest standard of privacy has been "the privacy of your own home." For example, the courts have ruled you to not have as much expectation of privacy in a car as you do in a home. Not as much on your open land as inside your house. Not as much in an RV as in a house. Your day-guests don't have the same privacy in your house as you do.

When technology changes privacy, there is always a battle over this expectation. The police want to reduce it so they can do more without warrants. The public wants to keep their privacy. Recently police made a number of arrests over the years by using infrared scanners to look for signals coming from inside a home. The court eventually told them no: That even though the IR was emitted from the house for view by "anybody" with the right equipment, what went on inside was still private. Often we lose privacy, as a man who lived in his motorhome discovered when the police were able to search it like a car because he was parked in a parking lot rather than an RV park.

Others have learned that what they say in front of baby monitors is not private, even though it is said inside their homes, because it is transmitted "in the clear" outside your home. (Cell phone calls, also often sent that way, have statutory protection of your privacy.)

E-mail privacy is in crisis because most e-mail is sent without encryption. That means that almost all e-mails are like postcards. Anybody with access to the wires (or airwaves) they travel over can read them. Because of this, arguments are being made that you should have no more expectation of privacy in an e-mail than you have in a postcard, or worse, a postcard you hand to a 3rd party to carry. This question will remain in the balance for some time.

GMail and its competitors may tip this balance, particularly if the e-mail managed by the webmail companies loses its ECPA protection status, the risk of which is described above. I send my e-mails from my own private machine, and that part is private. But I send them anywhere, including to webmail users where they are stored in a 3rd party database for searching. Of course, that database has password-controlled access and I would like to hope it's private, but there are those who will argue it's a step below the privacy expectation one has for one's own home computer. With a home computer, somebody has to break into your home or the computer to read the mail. On a 3rd party server, they can do that, but they might also see it due to a change of policy.

The fact that a webmail's privacy policies affect not just the presumably consenting user, but all their correspondents is of particular concern. And you don't know the target of your mail is using GMail. Many are forwarding their more normal looking address to GMail.

Sadly, we can also expect arguments that e-mails read from a database over an open wireless network might be akin to the cordless phone or baby monitor situation.

One can hope such arguments will not win. But what is true that the more e-mail that is sent unencrypted, the more it is stored in private but external databases, the greater the arguments will be that you no longer really can expect that your mail is secure and private from being seen by parties other than yourself and the recipients. If the courts ever become convinced this is true, e-mail could lose the protections it currently has, designed to parallel the protections of paper mail and phone calls.

My hope would be that Google's design should not reduce that expectation of privacy. But all e-mail vendors should work hard to ensure they don't even raise the risk, let alone pitch us over the edge. What a terrible thing if we were to lose the cherished expecatation of privacy we want in our letters in the name of convenience. The ECPA is just an ordinary law protecting e-mail privacy, not as strong as a constitutional protection.

The expectation of privacy issue is important because it has a bearing not just on users of GMail but on those who send mail to a GMail user. Since many people will alias other addresses into GMail, you may not be aware you are mailing GMail.

Search correlation

You've probably noticed that a lot of your web searches contain private information. I often type into Google things like the names of prescription drugs I have been given, to find out more about them. I use search engines to research the stocks I buy and look up my friends, or see what people are saying about my family. Very private stuff.

Largely we treat our searching as anonymous, though in fact it's not nearly so anonymous as we might like. All the major search engines use a cookie that allows the engine to correlate together all the searches we do. If, like me, you have broadband, your computer's numeric internet address is also fixed -- either permanently like mine, or effectively permanently as it is for all those who don't turn their computer off or who have a network gateway box.

All of the companies providing e-mail and search together create a troubling risk that the private matters in our e-mail can be combined with the things we search for. It's no surprise that this potential is there. Search companies are all eager to find ways to improve the relevancy of their search results in order to please their users. It's what we want them to do. Learning things about you is one obvious technique to do that.

Again this is part of a trend toward creating a giant dossier of all your private information in a central place. Because users will demand more accurate search, this trend won't be stopped. However, companies can be encouraged to anonymize data they collect, making it hard or impossible to link back to the real person

Google has said they are not correlating search and e-mail (or their social network prototype called Orkut.) That's good, but for business and user interface reasons they are not likely to say this will always be true. The browser cookie system allows the e-mail and search systems to share information about your identity unless you go to extreme lengths to delete your cookies or use cookie-washing software. Even then, unless you are a dial-up user, your IP address is almost as good as the cookie, perhaps better. Of course, once you have a common login for e-mail and enhanced search services your activities will be fully linked, as they can be on Yahoo today. You can only wash this through the extreme step of using a web anonymizer which bounces your requests among cooperating servers that hide your address.

Global Search

The war on terrorism has rewritten the rules on what's possible in civil rights. Consider the idea that the government might come with a warrant or new law to an e-mail provider and say, "Search all your customer's e-mail to see if anybody was talking about planes and the World Trade Center before 8am on Sept 11."

Under traditional 4th amendment rules, such a broad fishing expedition would be unthinkable. But some people are finding it more thinkable. Some might even think it's a good idea, and proclaim they don't mind if this were done to their own mail.

Without large webmail archives, the idea wasn't even possible. Now it could be.

What can Google Do

Google is proud of its reputation as a good corporate citizen that tries to keep its users interests at heart. It has done this even against the wishes of its customers (the advertisers) by putting restrictions on ads that other sites wouldn't place. This has actually been a good business decision, but Google espouses a philosophy of not doing any "evil" while building their business. It's a good philosophy.

Encryption

The most obvious step Google could take would be to encrypt a user's e-mail, searching index and other associated data, so it can only be accessed using the user's password, and of course that password should not be stored when an e-mail session is over.

If need be, the mail can be held temporarily unencrypted before delivery to the user (because then it has ECPA protection) and thus indexed and tied to ads. Then it can be encrypted. Both the index and the mail contents must be encrypted so that they can't be read without your password. Police could get a wiretap warrant to watch you type in your password, so this is not as private as doing fully encrypted mail to your home PC, but as noted the warrant process defends your rights much more than the subpoena or contractual system does.

An overseas escrow system would allow recovery of your password if you lose it, and in the inevitable event of your death.

GMail could also encourage the use of encryption in sending mail, both by doing SMTP-over-TLS (a standard technique for encrypting mail as it moves from server to server) wherever possible for your incoming and outgoing mail, and also supporting standard e-mail encryption formats like S/MIME and PGP, as well as new opportunistic encryption systems. This would be a giant step forward.

Resist correlation with search

This should be resisted as long as possible. If business needs -- which means pleasing the customer -- demand it, ways should be developed so that it's hard to tie the correlated data back to a particular user. If need be, it might even be better to have an explicit login (with password) so that the correlation data is also encrypted and only available when the user is logged on. Though frankly, I use Google search hundreds of times a day -- I'm always logged on.

Think about whether you can make such correlation opt-in, requiring the user's agreement. You might be able to create more accurate search doing this, but if customers want that, is it too hard to add the step of getting them to knowingly agree to it?

Lobby to strengthen the ECPA

Google, Microsoft, AOL, Yahoo and others should join with us in pointing out that the ECPA is now outdated, and not providing enough legal protection for all the new technological innovations in e-mail delivery. Fight in court cases where the DoJ and others try to expand their powers to grab user e-mail.

All these companies could also push for similar and stronger protections in the other nations where they do business.

Continue to refine policies and educate the public

Much of what Google is doing with GMail is innovative and worthwhile. It would be ridiculous to see it banned, as Senator Figueroa would suggest. It's not a lot different in kind from all the other well established webmail services -- but this doesn't mean those services didn't also have issues. There is not at this time, for example, a reason to be afraid of GMail and not afraid of premium Yahoo mail. Within months, the competitors will also release products with large archives and other features to keep up with Google.

So Google needs to refine their policies and contracts. Any policy that can be changed to protect user privacy that does no business harm to the company's plans should be changed. Even ones which do impede possible business plans should be evaluated and some of them done.

What users can do

The hard truth is if you are concerned about your privacy, 3rd-party hosted web based e-mail may not be right for you. There are tools that will let you search the e-mail on your PC.

You can also be more careful about how cookies are used to track you on the web. Most browsers can help you control this. I recommend the absolutely free Mozilla browser which lets you easily turn on and off what cookies go to what sites. Google is a good site that lets you use it without cookies; not all sites are so kind.

The problem with privacy is that nobody cares about their privacy until after it's been violated. Only when Bill Gates discovered that his e-mail could be searched for a message about cutting off "Netscape's Air Supply" did he realize the danger in logging it. Many people only realize the danger in their records when they enter a lawsuit, or a custody battle. They all thought it would never happen to them.

If you are concerned there are a variety of resources, including tools for anonymous web surfing, fancy cookie-management and more.

Most importantly, tell companies that you care about your privacy. Because others, not yet finding it violated, are not giving this message even though they do care if they think about it.

On the appearance of surveillance

To close, let me add one other rule about privacy. It is not only important to have your privacy. It is important that you believe you have your privacy. If you even suspect that you are being watched, it changes your behaviour and you become less free as an individual.

Are you the same person at your mother's house for holiday dinner as you were your first year on your own at university? How much you think you are being watched affects your freedom. How much a society thinks it is being watched affects the freedom of the society.

The fear that computerized scanning of our e-mails (to display ads or filter out spam) will result in actual harm is largely baseless. But even irrational fears affect our freedom, and this should be considered in software design.