How to make child-porn blocks safe for the internet

Last week, an obscure quasi-non-govermental agency called the Internet Watch Foundation was catapaulted into the spotlight on the news that they'd classed an image from Wikipedia as child pornography and that as a result, the page was blocked for 95% of the internet users in Britain. That's because the IWF is charged with creating a secret blacklist of child-porn URLs that it distributes to subscribed ISPs, who then block those links for their users.

Many reasonable people believed that the classification of the record cover was an error in judgement. It is the cover of Virgin Killer, a 1976 heavy metal album by the Scorpions. The image is of "a naked prepubescent girl" with a star of broken glass obscuring her genitals. The model herself has reportedly said that she isn't put out by the image's existence. The encylopedic context in which the image was reproduced is clinical, not titillating. After a vigorous public debate, the IWF withdrew its block, issuing a confusing and contradictory statement to the effect that even though the image is truly child pornography, they won't block it, unless it's on a British server, in which case they might.

It may be that this was the first time the IWF ever made a mistake, but it seems more likely that the "police-trained operators" who are charged with reviewing potentially obscene material commit errors at the same rate that the rest of us fallible mortals do. But we won't ever know about it, because of the way that IWF's blacklist is handled. When an internet user at an IWF-subscribed ISP visits a blocked page, s/he is usually presented with a "404" error page ("Not Found") or a "403" error ("Forbidden"). There is no way to differentiate between a page that you can't see because IWF is censoring it and a page that you can't see because the server is overloaded or the because the page has been relocated.

The only reason we know about this block is because of how Wikipedia handles its own blacklist – a list of IP addresses that have been used recently in vandalism against the encyclopedia. Traffic to sites that are on IWF's blacklist is all channeled through a small number of proxies, so once the Wikipedia page was blocked, it appeared to Wikipedia that every visitor from the participating ISPs shared the same addresses. This broke Wikipedia, triggering a lot of detective work by some very clever web users, who eventually concluded that the culprit was the IWF's (mis)classification.

There's plenty of material that pretty much everyone will agree is exploitative child pornography – for example, material on sites that self-identify as child-porn servers run by and for paedophiles. But there's also a large amount of material that is much harder to classify: Nabokov's Lolita, the nude photography of Charles Dodgson (AKA Lewis Carroll, who shot chaste-but-naked photos of baby girls), and the recent critically acclaimed graphic novel Lost Girls by Alan Moore (or, indeed, the cover of the Scorpions' Virgin Killer). Some of the controversy arises through the uneven application of standards: communities run by and for gay teenagers are more apt to be classified as obscene for their frank sexual talk than a comparable forum run by and for their straight peers.

As the father of a baby girl, I'd welcome any proposal that reduced the overall availability of child pornography and that could be used to catch, prosecute and treat or imprison those who exploit children. I want my daughter to be safe. I might even be willing to put up with the occassional false positive if the system was effective overall.

But the IWF system is one that both works poorly and fails poorly. News accounts of police prosecutions of pedophile rings inevitably involve the infiltration of private file-sharing systems, password-protected sites, email lists and P2P networks – areas that are all invisible to the IWF and unaffected by its filters (for that matter, IWF's filters can be defeated by using any of literally millions of public proxies). The incremental benefit of an IWF is that it protects the public from being tricked into seeing child porn (this is a benefit, make no mistake, though a lesser one than preventing children from being exploited or preventing the gratification of paedophiles) and it may discourage particularly unresourceful paedophiles who don't think to try P2P or a private site to get their kicks.

If IWF is going to make children safer from exploitation, it must succeed in reducing the number of paedophiles who get access to exploitative material – not in merely reducing the amount of exploitative material there is on the web. If the same paedophiles can access the same material using a different URL or protocol, the reduction in harm to exploited children is negligible. I don't want a system that squeezes the air into the other end of the balloon – I want one that lets the air out of the balloon altogether.

But stipulate that there is some good in moving the air around – protecting adults and children alike from seeing the images inadvertently, or discouraging the very dumbest and least committed of the paedo cohort. Fair enough – but let's at least make sure that when the system fails (as all systems do), that it fails well, in a way that minimises collateral damage.

Contrast the workings of IWF's list with the workings of HMRC, who are charged with examining the parcels that enter the country, some of which contain obscene material. If you order a copy of Lost Girls from a comics shop in New York, HMRC might very well open it at the border, have a quick look at the contents, and decide that this is obscene material. At that point, you'll get notice (or possibly even a visit from the police) that your parcel has been seized. Sometimes, they'll even notify the comics shop in New York. If you disagree with the judgement, you can appeal it internally with the Customs authority, and if you don't get satisfaction, you can ask a judge to rule on the case (something very much like this happened in Canada shortly after Lost Girls was published – and after a review, Canada Customs reclassified Lost Girls, clearing it for import).

This system fails well. We can ask our law enforcement officials to err on the side of caution, knowing that their errors can be corrected by the addressee asking for a review. The rule of law and due process are excellent, well-understood tools for addressing the fallibility of human beings in authority. Indeed, as grave as the matter of distributing child pornography is, it's a lesser crime than actually producing child pornography – and yet, those accused of the graver offence are guaranteed the due process that is denied to those who are accused of merely attempting to view it online.

IWF's blacklist lacks this fundamental check on its own fallibility. What if, instead of "404: Not found," pages on the blacklist returned this message: Material blocked: this page ("TITLE") appears on the Internet Watch Foundation's blacklist of obscene material. IWF officer NAME made this notation, "REASON_FOR_BLOCKING." If you believe that this is an incorrect classification, you can appeal the decision by visiting the IWF's appeals page (link). There is no penalty for appealing this decision, and your name and other details will be kept confidential."

Additionally, the IWF could notify (where contact details existed) the sitemaster for any page that they add to the blacklist. This is a few hours' worth of additional programming, in order to add the same system of checks and balances that prevails in the real world. If we're going to give a private entity the power to decide what we can and can't know, shouldn't we at least have a minimal means to detect their handiwork where it occurs?