Crikey: Internet filters a success, if success = failure

Crikey logo

[This article was first published in Crikey yesterday. I’ve added some follow-up comments at the end.]

Let’s sing along with Senator Conroy! You’ve got to accentuate the positive / Eliminate the negative / Latch on to the affirmative…

[On Monday] our Minister for Broadband was “encouraged” that lab tests of ISP-level Internet filters showed “significant progress” since 2005, and The Australian had him declaring the trial a success. But if you actually dig into the full report [2.8MB PDF] things aren’t so rosy.

Yes, on average filters might be more accurate than three years ago and have less impact on Internet speeds — well, at least for the six filters actually tested of the 26 put forward. But it’s about them being not quite as crap as before.

The report pre-judges the results, saying filters show “high levels of successful blocking”. But even with “most” filters achieving over 92% success, that still means 1 in 13 naughty sites are not blocked.

Similarly, the “low levels” of overblocking (incorrectly blocking legitimate content) are, at best, still 1%. With more than a million registered domain names in Australia (a loose measure of “sites”) even a 1% false positive rate means 10,000 perfectly acceptable websites are blocked. That’s with the best product. Under ideal lab conditions. The least successful of the products tested was eight times worse.

One product only degraded Internet speeds by 2%, maybe, but it was 22% to 30% for three products, and more than 75% for two of them. That’s up to 75% off your internet speed, or your ISP having to build 75% more capacity — with the cost passed on to you.

This was, remember, in a test lab. Filters were tested against a pre-compiled list of fewer than 4000 web pages (URLs). How they handle the massive, rapidly-changing real Internet, and how that affects performance of a real ISP, are different questions.

As the report notes, ACMA wasn’t asked to look at the balance of costs and benefits for ISP-level filtering, or the implications for customers, or how easy it is to circumvent the filters (“Very easy,” some reckon).

And here’s the killer. None of the products could effectively filter instant messaging, streaming video, peer-to-peer file sharing like BitTorrent, newsgroups or newly-invented Internet protocols except by blocking them entirely. Let’s count them again. None.

As the report notes, “Where such protocols are used to carry legitimate traffic and are widely used by children for study and social interaction, ACMA regards the absence of a more targeted capability as a deficiency.” Vendors mentioned development efforts but, writes ACMA, “Such capabilities may become available in the next few years.” Yeah, maybe. Until then, kids, go for it.

Hit it, Bing! You’ve got to accentuate the positive…

And beyond Crikey

One Crikey commenter noted that my analysis was fairly restrained. Perhaps. But I’ve written plenty about this. Filters simply will not solve the problem because they’re so easy to circumvent.

In particular, in Angry geeks: “Don’t waste money on Internet filters” six months ago I wrote:

Real-world experience in everything from spam filters to the record industry’s futile attempts to stop copyright violations always shows that filters only block casual users. Professionals, the desperate or the persistent will always get through.

However if a politician demands a filter, pretty soon a shiny-suited salesman will appear, ready to sell him a box with “filter” written on the front. It’ll work — well enough for the demo, anyway.

“Look, Minister! Nice Minister. Watch the screen. See? Filter off, bad website is visible. Filter on, bad website gone. Filter off. Child in danger. Filter on. Child happy and safe. Filter off. Voter afraid and angry. Filter on. Voter relaxed and comfortable. Cheque now please.”

It is obscene that further taxpayer’s money is being spent on the next trial when this report already shows — clearly shows — that the filters are simply not up for the job. Their false positive rate is unacceptably high. They’re useless for anything other than standard web traffic — yet non-web traffic is precisely where material of real concern is likely to be distributed.

The only people who think filters are the answer to the actual question of “keeping children safe” are those who know nothing about how the Internet works, and who want the government to do their babysitting for them.

Of course the real question the filters answer is “What do we do with Family First Senator Steve Fielding?” But that’s another story…

16 Replies to “Crikey: Internet filters a success, if success = failure”

  1. I will gladly accept the false positives and speed issues of an internet filtering system on 2 conditions:

    1. True 20MB+ broadband across the nation both wired and wireless!
    2. Decent data rates and no data caps for those plans.

    When the government has that sorted out come back and talk to me about filtering then.

  2. I have had the experience of my own web site being banned by the filter used at work as being a sex site. I had to write an angry email to my IT support staff to have it removed.

    @wolf, would you be happy if one fo the false positives was Medicare? No one would be able to make a claim.

  3. @yewenyi: My point was there are bigger issues with the infrastructure here in Australia than needing to slow it sown even more 🙂

    I think even <1% false positives is not acceptable. Esp for say teenagers trying to work out all the crap that is going on in their bodies and not being able to reach out and find information online. I have search “delicate ” terms to find medical information, and I most certainly don’t want the site on breast cancer e.t.c blocked because the government decided that it might be bad.

  4. A few other flaws in the report, testing methods and software:

    PC level vs ISP level filtering table
    At ISP level, you will be able to log into ISP Account and change your level of filtering so no different from Administrators password on PC

    Tests where conducted in Tasmania, which because of high cost of traffic to mainland makes heavy use of caching and proxies. So this is not your typical mainland tier 3 ISP.

    The best performing filter could not handle https, so either free for all or block everything.

    “Strong M material” got moved from category 3, which should not be blocked to category 2 which should be blocked. So did the people conducting this test decided they they knew better than the Censorship Board? or did they move them to get better results? This was material classified as M and should in theory be available to all, being blocked because the testers didn’t like it or it was screwing with their results. If they moved 50 sites, the results are +/- 5% accuracy on both blocking success and false positives.

  5. The link to Bing Crosby singing Accentuate the Positive is broken. “This video is no longer available due to a copyright claim by a third party,” says YouTube.

    So, mysterious Third Party, where would you like me to point people if they want to listen to the song and get the joke? Or are the few dollars you make from “protecting” a poor digitisation of a scratchy print of a film showing a long-dead singer more important than being a living part of our culture?


    You really are afraid, aren’t you!

Comments are closed.