Skip to main contentSkip to navigationSkip to navigation
Online privacy
Companies that offer privacy-oriented services are under increasing pressure to cooperate in surveillance operations. Photograph: Dominic Lipinski/PA
Companies that offer privacy-oriented services are under increasing pressure to cooperate in surveillance operations. Photograph: Dominic Lipinski/PA

'Warrant canaries': a subtle hint that your email provider is compromised

This article is more than 8 years old

The business of protecting data is getting tougher by the day – but there are ways not to breach the law and still safeguard customer data

Since the Snowden revelations, the market for privacy-oriented services has only grown – indeed, it’s likely that it will keep growing. We’re not at peak surveillance, but we’re way past peak indifference to surveillance.

Running a privacy service comes with two distinct technical challenges: the cryptographic challenge of making messages secure in transit and at rest on your server; and the legal challenge of keeping your promises to your customers intact when a government wants to spy on them.

These two problems are intimately entwined, and so are their solutions.

The cost of getting it wrong is high. Take Lavabit, the privacy-oriented email provider used by Edward Snowden to communicate with journalists when he was planning his leak: a few months after the Snowden revelations, nearly two years ago, Lavabit mysteriously shut its doors, its website replaced with a message saying that owner Ladar Levinson took the action rather than “being complicit in crimes against the American people.”

Later, it emerged that the NSA had secretly demanded that Lavabit insert a “backdoor” into its system so that it could potentially spy on all of Levinson’s customers. Rather than betray their trust, Levinson folded up his business altogether (braving arrest threats) a process he likened to “putting a beloved pet to sleep.”

Shortly after, Silent Circle, a major Lavabit competitor, pre-emptively shut the doors on its email service, apparently believing it was next in line for an NSA order. Groklaw, an investigative news site, also shut down, worried that it would have to secretly turn over information on its sources.

Since 9/11, governments around the world have created spying legislation that allows them to demand that companies cooperate in surveillance operations while being sworn to perpetual secrecy on pain of criminal prosecution. In response, companies created dead man’s switches, or “warrant canaries.”

Here’s how those work: companies periodically publish “transparency reports,” listing the number court orders they’ve received and whether and how they’ve complied with them. The first of these reports has a line like “Number of secret, gag-ordered surveillance warrants: 0”. If a company receives a secret warrant, it omits the line from its next transparency report. Eagle-eyed watchers note the omission and conclude that the service is compromised and may no longer be trusted. The service folds.

The idea of warrant canaries is not to voluntarily go out of business: it’s to make business-destroying secret warrants useless. “Serve us with one of your secret warrants,” they imply, “and everyone you wish to spy upon will automatically stop using this service, making the whole thing pointless.” This only works if you are based in a territory where the government can’t compel you to lie (that is, to go on publishing a transparency report that puts the number of secret warrants at zero, even when it’s non-zero).

Many US constitutional scholars believe that while a government could gag a business in the name of security without falling afoul of the first amendment, ordering it to utter falsehoods would not pass muster. Other countries are less protective of free expression: in Australia warrant canaries are now themselves illegal. Warrant canaries also only work if they’re published before the first snooping order arrives (that’s why the Electronic Frontier Foundation published an all-zeroes transparency report months before launching its new Let’s Encrypt certificate authority service).

Warrant canaries are a kind of Ulysses pact – a contract where one party begins negotiations by limiting their own choices (named for Ulysses, who tied himself to the mast to prevent himself from jumping in the sea when he heard the sirens’ songs). But as Ulysses pacts go, it’s a weak one: when it comes down to it, how many business owners are willing to shut the doors rather than tell one teeny-tiny lie on their transparency reports?

This matters, because the spy agencies serving the warrants are asking themselves the same question. A warrant canary only prevents secret spying warrants if the agency believes the business owner is committed to publishing – if they think the owner might just go along to get along, they have every incentive to go through with it.

There’s another kind of secret spying: malware implantation. This is when a government body orders a company to send some of its customers a software update that includes a backdoor. For example, the Saudi government once convinced Research in Motion to backdoor Blackberry devices within its borders. In May, 2014, the anonymously maintained Truecrypt project mysteriously shut down, leaving behind a cryptic note (possibly with a Dan-Brown-esque secret message in it). Many believe that they shut down in response to a government demand to weaken some or all of the Truecrypt programs in the wild.

In the case of programs that run on user’s computers there’s “binary transparency.” When a program with binary transparency receives an update, it computes that update’s “hash” (a mathematical fingerprint) and sends it to a server maintained by a disinterested third party. It also checks the hashes of all the other updates that have been received by all the other versions of the program that have checked in. If it sees that it has got a special update, it refuses to install it and alerts the user.

This is much stronger, more effective Ulysses pact. If a spy agency knows that any attempt to implant malware on a user’s computer through a software update will both fail and raise an alarm, there is absolutely no reason even to try.

What’s more, if binary transparency becomes the norm, than any company that doesn’t implement it will become the go-to target of opportunity for spies – they will be under huge pressure to match their competition.

As with many questions of technology and regulation, the answer to the problem is a blend of economics, technology and law – but when it comes to malware implantation, all three are on privacy’s side.

Comments (…)

Sign in or create your Guardian account to join the discussion

Most viewed

Most viewed