February 22, 2022 - 1:27pm

Forensic linguistics is fast becoming the solvent of internet anonymity.

In the end, it was forensic linguistics that did for the Unabomber, Ted Kaczynski. The former university professor made the mistake of publishing his manifesto. Industrial Society and Its Future. Armed with a large enough a data set, an FBI team used the fingerprints of his verbal style in the same way that a forensics team would use real fingerprints.

One member of the same team went on to identify ‘the longest serving attorney in the US Department of Justice’ as a notorious anonymous poster. And JK Rowling was ‘unmasked’ as Robert Galbraith by the decision to publish a data set in the form of Galbraith novel, A Cuckoo’s Calling.

In recent years, as more and more computers have been added to the mix, the identities that can be winkled out has become uncanny. This week, no less a shadow than Q himself was rumbled.

“Q” is the supposed Trump White House operative with top secret “Q-level” security clearance at the heart of the QAnon movement, who posts gnomic messages, ‘Q Drops’, on far-flung Internet message boards.

Using text samples with more than 100,000 words written by Q, and at least 12,000 words by each of the 13 other candidates who they analysed, two independent teams said their computers had honed in on two prime suspects: Paul Furber and Ron Watkins.

The conclusion itself is perhaps the least interesting part. Many hours of both podcast and documentary have already been spilled on establishing that Furber wrote early-Q before being superseded by Watkins in around 2018.

Much more gripping is how easy it has become to perform this kind of detection, as high-end computing and machine learning systems enter the picture. In a way, Q is just a high-profile version of a general use case. If one particularly infamous message board user can be found, why not all of them?

The forces of Something Must Be Done have long yearned for a de-anonymised internet. Even leathery everyman Adrian Chiles is a fan. Just a few months ago, a British MP was stabbed 17 times by a deranged Islamist, who didn’t have so much as an Odyssee account. Somehow, this morphed into yet more calls for an end to internet anonymity.

Rather than ending internet anonymity by fiat, it may be easier to end it by X-Ray. The principles remain the same. First, find a data set of someone’s utterances — Twitter is normally a good start. Next, match it against a dataset of a mysterious internet anon. Assess the correlation.

After all, we already live in a world where ‘anonymous trolls’ are regularly brought before the courts under the Communications Act 2003. Often, these people are found through their own carelessness — as with the autistic woman who was jailed for twelve weeks for sending offensive messages to feminist campaigner Caroline Criado Perez.

Now, the new Online Safety Bill that is about to go through Parliament will make it an offence to ‘knowingly distribute seriously harmful misinformation’. The steppingstones to a total dragnet over the “chans”, some parts of reddit, and the lowest tiers of Twitter hell, are clear.

Don’t imagine this would only be applied in a forward-going way. Just as the advent of DNA sequencing led to a rash of convictions for historical cases, so too, the capacity to induct from your casual misuse of the Oxford comma will be fodder for historians, employers, and police ‘anti-hate’ squads, in equal measure. Be careful what you wrote.

Gavin Haynes is a journalist and former editor-at-large at Vice.