EVEN before Edward Snowden’s revelations, many people suspected that spooks like those at America’s National Security Agency could pry into their electronic lives. Most were not worried enough to do anything about it. But political dissidents in authoritarian countries, and a hypervigiliant few in the West, have long tried to outwit the NSA and its counterparts in other countries. One favourite tool is The Onion Router, or TOR, a system which disguises the provenance of communications by rerouting them cleverly around the web. The New Yorker, among other media firms, uses TOR to help whistleblowers submit information anonymously.
But TOR is not perfect. Careful analysis of message timing, or the sort of traffic analysis made possible by the NSA’s enormous dragnet, may allow the spooks to penetrate the fog and pin down a user’s identity, even if they cannot read the messages themselves. Indeed, rumours have long swirled that the intelligence agencies run TOR nodes of their own, the better to keep tabs on its users. Happily, more help for the security-conscious whistleblower may now be at hand, in the form of AdLeaks, a system proposed by Völker Roth of the Free University in Berlin, in a paper posted to the online repository ArXiv.
As its name suggests, Dr Roth’s system takes its inspiration from online advertising networks such as Doubleclick. When you visit a website that displays adverts, your browser typically downloads code that performs various tasks on your computer. Such code is often used to target ads—for instance, by plugging stores in your vicinity that sell something a product that you recently searched for. AdLeaks operates in a similar way, except that its code allows a user’s computer to send small packets of encrypted information back to AdLeaks.
The system is designed to allow whistleblowers to hide in plain sight. Assuming that AdLeaks becomes widely distributed, as Dr Roth hopes it will, most of the blizzard of packets it will generate will not contain useful information. But because all the packets the system sends will be encrypted—scrambled, that is, into a random-seeming jumble of characters—it will not be possible for anyone listening in to tell the empty packets from the tiny few that contain real information. And because many people will be generating AdLeaks packets, the vast majority of whom will not be whistleblowers, the simple fact of generating such packets will not be enough to attract suspicion. AdLeaks’ servers, meanwhile, grab all such submitted packets, combine them and then decrypt them in a mathematically process that allows the system to reconstruct leaked documents piece by tiny piece (more mathematically-inclined readers can read the paper for a fuller description).
So far so good. But there is a slight hitch in the scheme. Genuine whistleblowers require special software to chop up the documents they wish to pass on, and to transmit those chunks in place of ordinary, content-free AdLeaks packets. Dr Roth cannot simply offer that software for download, as the very act of downloading it at all could be enough to arouse suspicion. One idea is to include the program on the DVDs still occasionally distributed with computing magazines. But a more elegant one is to use the AdLeaks system itself, by transmitting the whistleblowing software to everyone who is served with adverts in small encrypted chunks. A small, separtely-distributed “bootstrapper” program would then decrypt and install the software. Such a bootstrapping program is small and simple enough that it could simply be printed periodically by newspapers, or even distributed as a QCR code. Since the spooks cannot (yet) easily tell who is transcribing code from a piece of newspaper, or track photos of QCR codes, they need never learn that a whistleblower has installed the software.
That is not to say that AdLeaks is foolproof. For one thing, it relies on web browsers having good quality random number generators to power the encryption phase. Presently, most do not. An adversary could attempt to undermine the system in other ways, too. Simply blocking the AdLeaks servers would be enough to thwart it, and might be something that any prudent company would do on its own corporate network. If the whistleblower is already attracting suspicion, then secretly installing monitoring software on his computer would also give the game away.
A more elegant attack on the system exploits the way that its cryptography works, by generating a sacrificial leak in the hopes of confirming whether someone who has already fallen under suspicion is, in fact, leaking secrets. AdLeaks combines packets before decrypting them. Decrypting any number of empty packets will produce nothing. Decrypting any number of empty packets plus one meaningful packet will spit out that meaningful packet. But decrypting two or more meaningful packets at the same time will irretrievably destroy the information contained in both, an event called a “collision”.
To exploit this, an attacker must first take control of an AdLeaks computer. He then creates a leak of his own (perhaps genuine, or perhaps fake), chops it up and encrypts it. He combines his own meaningful packets with those generated by the person under suspicion, and submits the results to AdLeaks. Then it is merely a matter of keeping an eye on the newspapers. If his fake leak appears in the press, that means it survived the combination process. That, in turn, means that the suspected whistleblower was actually transmitting empty packets and is probably innocent. But if the story fails to appear, then the suspect may be guilty—his meaningful packets may have collided with the attackers, leaving nothing to pass on to AdLeaks’ customers in the press. (Though there are admittedly other reasons why a whistle fails to blow; newspapers need not publish every scoop they get their hands on.)
All this is technically nifty. But the cost of setting up and running the servers for such a system could be substantial. Dr Roth calculates that a system that served ads to 138m internet surfers would cost around $600 a day—a significant expense for the newspapers, NGOs and underground political movements likely to benefit from the system. But that may not matter. For AdLeaks is a whistleblowing system wrapped up in the entirely legitimate—and profitable—business of web advertising. On the fairly conservative assumption that AdLeaks were to pay out $0.25 per thousand impressions (currently the going rate for a fairly run-of-the-mill website) and to charge a modest markup, then the business might even cover its costs.
Powered by Facebook Comments