For months now Communications Minister Stephen Conroy has been answering letters about his internet censorship drive with generalities about "cyber-safety". He has deferred any further debate on the policy until after the upcoming "live" filtering trial. Just this week, he deflected comments about the scope of the proposed blacklist in a Senate Estimates hearing, saying "it will be determined, as we have always said, following the live pilot trial."
Testing is expected to begin in earnest in the coming weeks and the trial’s progress is being watched keenly by those who have an interest in how the filtering policy will take shape.
The pilot, according to the "expression of interest" document released by the Government, is designed to "test a range of content filtering solutions in a real world environment". The outcomes, it says, will "inform the Government’s decision making".
Data will be gathered on the accuracy, costs, scalability and effect on network speed brought about by the filtering software, as well as commercial concerns.
Even before the trial has begun, questions are being asked about its integrity. The process by which participating ISPs were selected is a little opaque. ISPs were invited to volunteer to take part in the pilot, and several did so. The ultimate decision, however, to exclude the nation’s most popular ISPs has raised a few eyebrows.
Although the nation’s largest ISP, Telstra, did not choose to take part in the trial, the second- and third-largest, Optus and iiNet, did apply. Neither were selected. Instead, six smaller ISPs were named earlier this month to take part in the first round of testing: Primus Telecommunications, Tech 2U, Webshield, OMNIconnect, Netforce and Highway 1.
These ISPs represent a very small percentage of Australian internet users. In fact, one of them, Highway 1, is a business-focussed ISP and has hardly any home users at all.
Tests conducted on the networks of some of the country’s smallest ISPs cannot possibly reflect the realities and challenges of filtering on a national level, where millions of users are viewing thousands of web pages every second.
Presumably, this decision was partly based on cost: outfitting the larger networks of iiNet or Optus, even for a limited trial, would be very expensive. Smaller ISPs are probably also more able to respond in the trial’s very short time-frame. Nevertheless, the credibility of the test is undermined considerably by its limited scale.
Cynics have also suggested that iiNet was excluded because of their vocal opposition to the scheme as a whole. The company’s chief executive has been quoted as saying the company applied to take part in the trial merely to show "how stupid it is".
Not all the ISPs are so pessimistic. Some maintain it is technically feasible and see it as a revenue opportunity. Nicholas Power, General Manager of Highway 1, told tech journalist Ben Grubb of Tech Wired AU that participating in the pilot made "good business sense" — by being closely involved with the development of the policy, they could use Government money to prepare themselves, should the legislation come into effect.
Those chosen to take part must undertake to filter a blacklist of websites based on the ACMA list of prohibited material. This simulates the mandatory tier of the filtering policy targeted at illegal and "undesirable" material.
The choice of technology used to filter content is up to the ISPs, and they can apply for government funding to offset the cost of the new equipment. The ISPs must then provide a filtered internet connection to Enex Testlabs, who will be conducting the test and eventually delivering a report to the Government. The connection will be tested for 60 days meaning that, barring leaks, the public is unlikely to hear much about the trial until May.
We know now what the pilot is testing, but it is what the pilot will not test that is problematic. Most noticeably missing from the pilot is the policy that was actually taken to the last federal election: a child-friendly internet feed. The "technical testing framework" document mentions "dynamic analysis filtering" in passing, but the focus is clearly on the mandatory tier of the scheme. The ACMA blacklist, although reasonably broad in scope, is small and has been presented by the Minister as a tool for fighting the spread of child abuse material by pedophiles, not for protecting underage web surfers. Filtering just the ACMA blacklist is certainly more technically realisable than dynamic filtering of all requested content, but it means the trial will not reflect the Government’s proposed policy in full.
Exactly how the various ISPs will approach the trial will be interesting to see. Filtering the blacklist can be accomplished relatively simply at a network level if blunt instruments are used — that is, if whole domain names are blocked rather than particular pages. It is not yet clear, however, how the overblocking which can arise through DNS poisoning and IP blocking will be avoided.
A blocking solution that can filter out individual web pages in a site is much more complicated technically. We saw last year how necessary such a system is when the Internet Watch Foundation, which compiles an index of child-abuse URLs for blocking purposes, added a Wikipedia page to their list. Blocking the entire domain would be catastrophic. The same applies for many sites such as YouTube that host millions of pages of content, only one of which needs to be objectionable to complicate access to the entire domain.
There has been some confusion as to whether or not the pilot will even include actual users. Initially, the Government indicated that "this will be a closed network test and will not involve actual customers". Participating ISPs, however, have said they will offer filtering to their customers. Whether or not the experiences of end users will inform the final report is not yet known.
The ISPs are currently in the process of configuring the hardware
and software they have chosen to implement the filtering. They will
have to put this to the Government for approval before
implementation, which indicates the Department is keeping a close eye
on the technology on offer. This makes good sense, as if and when the
system is rolled out nationwide, the Government will have to subsidise
the cost, which could run to tens of millions of dollars.
The Government knows filters can be trivially broken and points this out several times in its own documents: "It is acknowledged that filtering can be circumvented by motivated people with sufficient technical knowledge." If the pilot does show, as the Government itself seems to anticipate, that filter circumvention is trivial for most adults, how will this impact the proposed policy?
Regardless of the outcome of this trial, having some concrete data to focus the debate should be welcomed by all sides. Some hard technical data and real cost figures will allow a proper cost-benefit analysis to finally be undertaken.
That said, given its rather limited scope, it is disingenuous for the Minister to hide behind the trial and refuse to engage his critics in the meantime. Fundamental policy questions remain — such as what online content is to be effectively outlawed by the filter. A small ISP, let alone a large one, cannot inform this debate through the purchase of a new router. As Tech 2U’s General Manager Andrew Robson points out, they can test the network, but the question of where parental responsibility ends and Government mandate begins has not yet been satisfactorily discussed or resolved.
For the technical numbers to have a meaningful context, these issues must be clarified.