So how does Echelon work? At first glance, United States Patent 5,937,422 looks like a fairly innocuous description of an extremely abstract process. On further examination, however, the patent registered to assignee "The United States of America as represented by the National Security", filed April 15th of 1997, takes on some sinister overtones, the least of which are shaded by the identity of the assignee.

Briefly, the technology works like so: given text input, such as standard e-mail, the output from a voice-to-text converter, an HTML posting or a message on a Usenet newsgroup, it is capable of sorting words it searches for in a hierarchical structure, categorizing and arbitrarily prioritizing the words based on a complex value assignment system.

The practical application? Well, perhaps some sort of advanced logging utility that's responsible for interpreting, parsing, and prioritising the output of monitored processes (like you might see in an intrusion detection system). Perhaps a marketing genius could construct an abstract algorithm - "monitoring media and public communications to predict the likeliness of certain new trends arising" - using some of this technology. Why, it even produces a 'Topical Description' to help clarify the output!

However, we have to ask ourselves what the National Security Agency is likely to do with such tools, as the patent seems to belong to them. If you were in the business of intelligence, what would your priorities be?

Generally, intelligence work does not consist of James Bond style sleuthing - very little cloak and dagger. In fact, it mostly consists of gathering garbage - massive, cheap, readily available snippets of data - and then diligently analysing the resulting haystack for the glint of a needle. But let's say that you suddenly had a tool at your disposal: a rough but functional sorting machine, that could very reliably tear through the hay and pick out the needles. Naturally, it would take a willing attendant to ensure that needles were not being missed - but would it not make the entire process in many ways more economical, more manageable, and less of a usually-fruitless hassle?

As described by many sources, the bounty of Echelon lies in its automated sorting and prioritizing abilities. You could hire thousands upon thousands of potential security risks to tap millions of lines, or you could hire a few hundred well trained agents, develop some technology, and look only for the needles that may have been missed.

An entrepreneurial management might even opt for a buy in - you help us snag more hay, and we'll allow you to supply some of the criterion by which the needles are found. And this, it seems, has actually happened. With the roster of countries participating in Echelon, at what seem to be 'Buy-In' and 'Partner' levels, we can't help but speculate at the possibility of an extremely lucrative tradeoff. Knowledge is still power, after all.

'Biomedical engineers at the University of Southern California have created the world's first machine system that can recognize spoken words better than humans can' - Navy Wire Service, November 1, 1999.

The rough technology has been around for quite awhile, and it's unlikely that many have failed to notice the recent advances in speech recognition software. Of course, the allegations that Echelon was tapping undersea cables and modern-day POTS networks fell like rain prior to 1999, but this technological advancement has essentially resulted in fewer stalks of hay in the needle bin.

As the technology gets cleaner, fewer matches are the result of errors in translation. And, there are sources of intelligence to consider that don't even require this somewhat impressive technology.

I can personally prove, from my work with Internet engineering, that monitoring public traffic on a network (such as the Internet) can easily be engineered with no noticable interruption in service. The very nature of the Internet is to hide and invalidate isolated and unconnected service interruptions - but many seem to believe that something drastic and noticable must occur in order for such implementations to occur. A program known as a 'packet sniffer' is a readily available tool which allows the contents of all passing traffic to be examined. In many cases this only pertains to 'local' traffic, but not in all. Using just tools built in to the UNIX operating system, you could create and maintain your own Internet surveillance post, sorting through the vast amount of data captured for whatever sort of information you're looking for. Naturally, it would require a bit of timing and influence to implement such stations in the right places to catch the majority of Internet transactions - but it's not all that complicated. And if anybody's got timing and influence, it's our friends at the NSA.

Telephone networks are equally suitable to surreptitious monitoring. Rerouting is commonly performed for the purposes of maintenance and equipment upgrades. It wouldn't be difficult to install 'supervisor' measures to permit the duplication and availability of information passing over the lines in a given piece of equipment during one of these commonplace maintenance windows.

The technology for these terrifying prospects is not new - the means for interception have stepped hand in hand down the aisle with the means for communication all along.

Frances Farmer [e-mail]

See also:
Echelon (I) History and Politics

Sources:
US Patent 5,937,422
[Part Two]

Discuss this article: Underground: Switchboard